Meet the 2023 Innovators Under 35 Eric Schmidt on transforming science When AI goes to war Volume 126 Number 5 September/October 2023 Experimental drugs The ethics issue Who should get them? SO23-front_cover.indd 1 8/1/23 8:49 AM The Tim e ClimateTech SO23.CT.spread.final.indd 2 Join us on campus October 4-5, 2023 ClimateTechMIT.com 7/20/23 9:23 AM m e Is Now Innovations for a sustainable future ClimateTech convenes the leaders funding, creating, and deploying the technologies to lead the transition to a green economy. SO23.CT.spread.final.indd 3 7/20/23 9:23 AM From the editor In his essay introducing this year’s class of Innovators Under 35, Andrew Ng argues that AI is a general-purpose technology, much like electricity, that will be built into everything else (page 74). Indeed, it’s true, and it’s already happening. AI is rapidly becoming a tool that powers all sorts of other tools, a technological underpinning for a range of applications and devices. It can helpfully suggest a paella recipe in a web app. It can predict a protein structure from an amino acid sequence. It can paint. It can drive a car. It can relentlessly replicate itself, hijack the electrical grid for unlimited processing power, and wipe out all life on Earth. Okay, so that last one is just a nightmare scenario courtesy of the AI pioneer Geoffrey Hinton, who posed it at an EmTech Digital event of ours earlier this year. But it speaks to another of Ng’s points, and to the theme of this issue. Ng challenges the innovators to take responsibility for their work; he writes, “As we focus on AI as a driver of valuable innovation throughout society, social responsibility is more important than ever.” In many ways, the young innovators we celebrate in this issue exemplify the ways we can build ethical thinking into technology development. That is certainly true for our Innovator of the Year, Sharon Li, who is working to make AI applications safer by causing them to abstain from acting when faced with something they have not been trained on (page 76). This could help prevent the AIs we build from taking all sorts of unexpected turns, and causing untold harms. This issue revolves around questions of ethics and how they can be addressed, understood, or intermediated through technology. Should relatively affluent Westerners have stopped lending money to small entrepreneurs in the developing world because the lending platform is highly compensating its top executives? How much control should we have over what we give away? These are just a few of the thorny questions Mara Kardas-Nelson explores about a lenders’ revolt against the microfinance nonprofit Kiva (page 38). On page 24, Jessica Hamzelou interrogates the policies on access to experimental medical treatments that are sometimes a last resort for desperate patients and their families. Who should be able to use these unproven treatments, and what proofs of efficacy and (more important) safety should be required? In another life-and-death question, Arthur Holland Michel takes on computer-assisted warfare (page 46). How much should we base our lethal decision-making on analysis performed by artificial intelligence? How can we build those AI systems so that we are more likely to treat them as advisors than deciders? Rebecca Ackermann takes a look at the long evolution of the open-source movement (page 62) and the ways it has redefined freedom—free as in beer, free as in speech, free as in puppies— again and again. If open source is to be something we all benefit from, and indeed that many even profit from, how should we think about its upkeep and advancement? Who should be responsible for it? SO23-front_editorial.indd 2 Mat Honan is editor in chief of MIT Technology Review And on a more meta level, Gregory Epstein, a humanist chaplain at MIT and the president of Harvard’s organization of chaplains, who focuses on the intersection of technology and ethics, takes a deep look at All Tech Is Human, a nonprofit that promotes ethics and responsibility in tech (page 32). He wonders how its relationship with the technology industry should be defined as it grows and takes funding from giant corporations and multibillionaires. How can a group dedicated to openness and transparency, he asks, coexist with members and even leaders committed to tech secrecy? There is a lot more as well. I hope this issue makes you think, and gives you lots of ideas about the future. Thanks for reading, Mat Honan ROBYN KESSLER 02 7/29/23 8:21 AM Syngenta and Infosys: 20 years of relentless collaboration for shared success. www.technologyreview.com/thecloudhub Untitled-2 1 7/31/23 13:23 04 Contents “We might not have the opportunity to wait to take one of those other drugs that might be made available years down the line.” –p. 24 24 The right to try Cover story: Desperate people will often want to try experimental, unproven treatments. How can we ensure they’re not exploited or put at risk? BY JESSICA HAMZELOU 32 Only human Tech culture is increasingly oriented around moral and ethical messages: So why not a tech ethics congregation? BY GREG M . EPSTEIN Front 2 Letter from the editor THE DOWNLOAD 9 Eric Schmidt on how AI will transform science; better weather predictions; fawning over the Frequency Allocation Chart; extracting climate records from Antarctic ice cores; and saving Venice from sinking. Plus, job of the future: chief heat officer EXPLAINED 18 Everything you need to know about the wild world of alternative jet fuels How french fries, trash, and sunlight could power your future flights. By Casey Crownhart PROFILE 20 Valley of the misfit tech workers Xiaowei Wang and Collective Action School seek to remedy the moral blindness of Big Tech. By Patrick Sisson SO23-front_contents.indd 4 38 What happened to Kiva? Hundreds of lenders are protesting changes at the microfinance funder. Is their strike really about Kiva, or about how much control we should expect over international aid? BY MARA KARDAS - NELSON 46 AI-assisted warfare If a machine tells you when to pull the trigger, who is ultimately responsible? BY ARTHUR HOLLAND MICHEL Back 54 The greatest slideshow on Earth From supersize slideshows to Steve Jobs’s Apple keynote, corporate presentations have always pushed technology forward. By Claire L. Evans 62 Open source at 40 Free and open-source software are now foundational to modern code, but much about them is still in flux. By Rebecca Ackermann 68 Tiny faux organs could finally crack the mystery of menstruation Organoids are helping researchers explore one of the last frontiers of human physiology. By Saima Sidik 74 35 Innovators Under 35 Tips for aspiring innovators on trying, failing, and the future of AI. By Andrew Ng 76 Innovator of the Year: Sharon Li By Melissa Heikkilä 78 Online fraud, hacks, and scams, oh my Three books that explore how culture drives foul play on the internet. By Rebecca Ackermann FIELD NOTES 84 Servers that work from home Wasted heat from computers is transformed into free hot water for housing. By Luigi Avantaggiato ARCHIVE 88 A cell that does it all For 25 years, embryonic stem cells have been promising and controversial in equal measure. How far have they really come? COVER ILLUSTRATION BY SELMAN DESIGN The ethics issue 8/1/23 3:40 PM Discover what’s coming next in technology. Subscribe now for access to: • • • • • In depth reporting on AI, climate change, biotech & more Trusted insights you can’t find anywhere else Science & technology news shaping the future 6 print and digital issues a year Discounts on MIT Technology Review events Scan this code to subscribe or learn more at technologyreview.com/subscribe Untitled-1 1 6/8/23 2:59 PM 06 Masthead Editorial Corporate Consumer marketing Editor in chief Chief executive officer and publisher Mat Honan Elizabeth Bramson-Boudreau Vice president, marketing and consumer revenue Executive editor, operations Amy Nordrum Executive editor, newsroom Niall Firth Editorial director, print Allison Arieff Editorial director, audio and live journalism Jennifer Strong Editor at large David Rotman Science editor Mary Beth Griggs News editor Charlotte Jee Features and investigations editor Amanda Silverman Managing editor Timothy Maher Commissioning editor Rachel Courtland Senior editor, MIT News Alice Dragoon Senior editor, biomedicine Antonio Regalado Senior editor, climate and energy James Temple Senior editor, AI Will Douglas Heaven Podcast producer Anthony Green Senior reporters Eileen Guo (features and investigations) Jessica Hamzelou (biomedicine) Melissa Heikkilä (AI) Tate Ryan-Mosley (tech policy) Finance and operations Chief financial officer, head of operations Enejda Xheblati General ledger manager Olivia Male Accountant Anduela Tabaku Human resources director Alyssa Rousseau Manager of information technology Colby Wheeler Data analytics manager Audience engagement editor Juliet Beauchamp Creative director, print Eric Mongeon Digital visuals editor Stephanie Arnett Alliya Samhat Director of event marketing Nina Mehta Email marketing manager Tuong-Chau Cai Circulation and print production manager Tim Borton Advertising sales Senior manager of licensing Ted Hu Senior editor, custom content Michelle Brosnahan Senior editor, custom content Kwee Chuan Yeo Editor, custom content Teresa Elsey Senior project manager Martha Leibs Director of partnerships, Europe Technology Associate vice president, integrated marketing and brand Chief technology officer Drake Martinet Caitlin Bergmann caitlin.bergmann@technologyreview.com Vice president, product Executive director, brand partnerships Mariya Sitnova Molly Frey Marii Sebahar marii@technologyreview.com 415-416-9140 Associate product manager Executive director, brand partnerships Senior software engineer Allison Chase Digital brand designer Vichhika Tep Events Senior vice president, events and strategic partnerships Amy Lammers Director of event content and experiences Brian Bryson Senior event content producer Kristin Ingram kristin.ingram@technologyreview.com 415-509-1910 Executive director, brand partnerships Stephanie Clement stephanie.clement@ technologyreview.com 214-339-6115 Natasha Conteh Emily Kutchinsky Director of partnerships, Asia Marcus Ulvne Board of directors Cynthia Barnhart, Cochair Alan Spoon, Cochair Lara Boro Peter J. Caruso II, Esq. Whitney Espich Sanjay E. Sarma David Schmittlein Glen Shor Customer service and subscription inquiries National 877-479-6505 Executive director, sales and brand partnerships International Debbie Hanley debbie.hanley@technologyreview.com 214-282-2727 Email Senior director, brand partnerships 847-559-7313 customer-service@technologyreview.com Web www.technologyreview.com/ customerservice Director of events Ian Keller ian.keller@technologyreview.com 203-858-3396 Nicole Silva Senior director, brand partnerships techreview@wrightsmedia.com 877-652-5295 Erin Underwood Event operations manager Elana Wilner Manager of strategic partnerships Madeleine Frasca Williams Event coordinator Bo Richardson Miles Weiner miles.weiner@technologyreview.com 617-475-8078 Senior director, digital strategy, planning, and ad ops Katie Payne katie.payne@technologyreview.com Digital operations coordinator Brooke McGowan brooke.mcgowan@technologyreview.com Media kit www.technologyreview.com/media SO23-front_masthead.indd 6 Laurel Ruma Linda Cardinal Office manager Marcy Rizzo Abby Ivory-Ganja Director of acquisition marketing Global director of custom content Project manager Casey Crownhart (climate and energy) Rhiannon Williams (news) Zeyi Yang (China and East Asia) Senior audience engagement editor Taylor Puskaric Nicola Crepaldi Andrew Hendler andrew.hendler@technologyreview.com 201-993-8794 Christopher Doumas Head of international and custom events Linda Lowenthal Director of retention marketing Vice president, Insights and international Senior vice president, sales and brand partnerships Reporters Copy chief Alison Papalia MIT Technology Review Insights and international Reprints Licensing and permissions licensing@technologyreview.com MIT Technology Review 196 Broadway, 3rd Floor Cambridge, MA 02139 617-475-8000 Our in-depth reporting reveals what’s going on now to prepare you for what’s coming next. Technology Review, Inc., is an independent nonprofit 501(c)(3) corporation wholly owned by MIT; the views expressed in our publications and at our events are not always shared by the Institute. 7/28/23 10:25 AM Meet this year’s 35 Innovators Under 35 Become an MIT Technology Review digital subscriber and be among the first to meet the young innovators, leaders, and entrepreneurs shaping the future of technology. With your digital subscription, you’ll get full access to the list of honorees and their stories as soon as they’re published online on September 12th. Scan this code to become a digital subscriber or visit TechnologyReview.com/SubTR35 Untitled-5 1 8/3/23 10:59 Join us online or on the MIT campus | Coming November 2023 MIT Technology Review’s flagship event brings together global change makers, innovators, and the leading voices in industry to understand the impact and set a new course forward. EmTechMIT.com SO23.ETM.single.final.indd 1 PHOTO: MUZAMMIL SOORMA AI, climate change, and biotech are reshaping our economy and our lives 7/20/23 9:13 AM 09 The Download This is how AI will transform the way science gets done Science is about to become much more exciting—and that will affect us all, argues Google’s former CEO. By Eric Schmidt Reimagining science At its core, the scientific process will remain the same: conduct background research, identify a hypothesis, test it through experimentation, analyze the collected data, and reach a conclusion. But AI has the potential to revolutionize how each of these components looks in the future. COURTESY PHOTO With the advent of AI, science is about to become much more exciting—and in some ways unrecognizable. The reverberations of this shift will be felt far outside the lab and will affect us all. If we play our cards right with sensible regulation and proper support for innovative uses of AI to address science’s most pressing issues, it can rewrite the scientific process. We can build a future where AI-powered tools will both save us from mindless and time-consuming labor and encourage creative breakthroughs that would otherwise take decades. AI in recent months has become almost synonymous with large language models, or LLMs, but in science there are a multitude of different model architectures that may have even bigger impacts. In the past decade, most progress in science has come through smaller, “classical” models focused on specific questions. These models have already brought about profound advances. More recently, larger deep-learning models that are beginning to incorporate cross-domain knowledge and generative AI have expanded what is possible. Scientists at McMaster and MIT, for example, used AI to identify an antibiotic that fights what the World Health Organization calls one of the world’s most dangerous drug-resistant bacteria for hospital patients. The FDA has already cleared 523 devices that use AI, and a Google DeepMind model can control plasma in nuclear fusion reactions, bringing us closer to a clean-energy revolution. SO23-front_thedownload.indd 9 7/31/23 10:58 AM 10 The Download Starting with the research step, tools like PaperQA and Elicit harness LLMs to scan databases of articles and produce succinct and accurate summaries of the existing literature— citations included. Next, AI can spread the search net for hypotheses wider and narrow the net more quickly. As a result, AI tools can help formulate stronger hypotheses, such as models that spit out more promising candidates for new drugs. We’re already seeing simulations running multiple orders of magnitude faster than just a few years ago, allowing scientists to try more design options in simulation before carrying out real-world experiments. Moving on to the experimentation step, AI will be able to conduct experiments faster, cheaper, and at greater scale. Instead of limiting themselves to just six experiments, scientists can use AI tools to run a thousand. Scientists who are worried about their next grant, publication, or tenure process will no longer be bound to safe experiments with the highest odds of success, instead free to pursue bolder and more interdisciplinary hypotheses. Eventually, much of science will be conducted at “self-driving labs”—automated robotic platforms combined with artificial intelligence, which are already emerging at organizations like Emerald Cloud Lab, Artificial, and even Argonne National Laboratory. Finally, at the stage of analysis and conclusion, self-driving labs will move beyond automation and use LLMs to interpret experimental results and recommend the next experiment to run. The AI lab assistant could then order supplies and run that next recommended experiment overnight—all while the experimenter is home sleeping. Young researchers might be shifting nervously in their seats at this prospect. Luckily, the new jobs that emerge from this revolution are likely to be more creative and less mindless than most current lab work. With LLMs able to assist in building code, STEM students will no longer have to master obscure coding languages, opening the doors of the ivory tower to new, nontraditional talent and allowing scientists to engage with fields beyond their own. Soon, specifically trained LLMs might be developed to offer “peer” reviews of new papers alongside human reviewers. We must nevertheless recognize where the human touch is still important and avoid running before we can walk. For example, a lot of the tacit knowledge that scientists learn in labs is difficult to pass on to AI-powered self-driving labs. Similarly, we should be cognizant of the limitations of current LLMs—such as limited memory and even hallucinations—before we offload much of our paperwork, research, and analysis to them. The importance of regulation AI is such a powerful tool because it allows humans to accomplish more with less: less time, less education, less equipment. But these capabilities make it a dangerous weapon in the wrong hands. University of Rochester professor Andrew White was contracted by OpenAI to participate in a “red team” that could expose the risks posed by GPT-4 before it was released. Using the language model and giving it access to tools, White found it could propose dangerous compounds and even order them from a chemical supplier. To test the process, he had a (safe) Weather forecasting is having an AI moment As extreme weather conditions become more common, accurate forecasts become even more important. By Melissa Heikkilä The first week of July was the hottest week on record—yet another sign that climate change is “out of control,” the UN secretary general said. Punishing heat waves and extreme weather events like hurricanes and floods are going to become more common as the climate crisis worsens, making it more important than ever before to produce accurate weather forecasts. SO23-front_thedownload.indd 10 AI is proving increasingly helpful with that. In the past year, weather forecasting has been having an AI moment. Three recent papers from Nvidia, Google DeepMind, and Huawei have introduced machine-learning methods that are able to predict weather at least as accurately as conventional methods, and much more quickly. Recently I wrote about Pangu-Weather, an AI model developed by Huawei. Pangu-Weather is able to forecast not only weather but also the path of tropical cyclones. Huawei’s Pangu-Weather, Nvidia’s FourcastNet, and Google DeepMind’s GraphCast are making meteorologists “reconsider how we use machine learning and weather forecasts,” Peter Dueben, head of Earth system modeling at the European Centre for Medium-Range Weather Forecasts (ECMWF), told me for the story. ECMWF’s forecasting model is considered the gold standard for medium-term weather forecasting (up to 15 days ahead). Pangu-Weather managed to get accuracy comparable to that of the ECMWF model, while Google DeepMind claims in a nonpeer-reviewed paper to have beaten it 7/31/23 10:58 AM The Download compound shipped to his house the next week. OpenAI says it used his findings to tweak GPT-4 before it was released. OpenAI has managed to implement an impressive array of safeguards, but the day will likely soon come when someone We need smart, well-informed regulation—on both tech giants and open-source models—that doesn’t keep us from using AI in ways that can be beneficial to science. manages to copy the model and house it on their own servers. Such frontier models need to be protected to prevent thieves from removing the AI safety guardrails so carefully added by their original developers. To address bad uses of AI, we need smart, well-informed regulation—on both tech giants and open-source models—that doesn’t keep us from using AI in ways that can be beneficial to science. Beyond regulation, governments and philanthropy can support scientific projects with a high social return but little financial return or academic incentive, such as those in climate change, biosecurity, and pandemic preparedness. Insofar as safety concerns allow, government can also help develop large, high-quality data sets such as those that enabled AlphaFold, the model developed by Google’s DeepMind that predicts a protein’s shape from a sequence of amino acids. Open data sets are public goods: they benefit many researchers, but 90% of the time in the combinations that were tested. Using AI to predict weather has a big advantage: it’s fast. Traditional forecasting models are big, complex computer algorithms based on atmospheric physics, and they take hours to run. AI models can create forecasts in just seconds. But they are unlikely to replace conventional weather prediction models anytime soon. AI-powered forecasting models are trained on historical weather data that goes back decades, which means they are great at predicting events that are similar to the weather of the past. That’s a problem in an era of increasingly unpredictable conditions. We don’t know if AI models will be able to predict rare and extreme weather events, says Dueben. He thinks the way SO23-front_thedownload.indd 11 11 researchers have little incentive to create them themselves. Chemistry, for example, has one language that unites the field, which would seem to lend itself to easy analysis by AI models. But no one has properly aggregated data on molecular properties stored across dozens of databases, which keeps us from accessing insights into the field that would be within reach of AI models if we had a single source. Biology, meanwhile, lacks the known and calculable data that underlies physics or chemistry, with subfields like intrinsically disordered proteins still a mystery to us. It will therefore require a more concerted effort to understand—and even record—the data for an aggregated database. The road ahead to broad AI adoption in the sciences is long, with a lot that we must get right, from building the right databases to implementing the right regulations; from mitigating biases in AI algorithms to ensuring equal access to computing resources across borders. Nevertheless, this is a profoundly optimistic moment. Previous paradigm shifts in science, like the emergence of the scientific process or big data, have been inwardly focused—making science more precise, accurate, and methodical. AI, meanwhile, is expansive, allowing us to combine information in novel ways and to bring creativity and progress in the sciences to new heights. Eric Schmidt was the CEO of Google from 2001 to 2011. He is currently cofounder of Schmidt Futures, a philanthropic initiative that brings talented people together in networks to prove out their ideas and solve hard problems in science and society. forward might be for AI tools to be adopted alongside traditional weather forecasting models to get the most accurate predictions. Big Tech’s arrival on the weather forecasting scene is not purely based on scientific curiosity, reckons Oliver Fuhrer, the head of the numerical prediction department at MeteoSwiss, the Swiss Federal Office of Meteorology and Climatology. Our economies are becoming increasingly dependent on weather, especially with the rise of renewable energy, says Fuhrer. Tech companies’ businesses are also linked to weather, he adds, pointing to everything from logistics to the number of search queries for ice cream. The field of weather forecasting could gain a lot from the addition of AI. Countries track and record weather data, Using AI to predict weather has a big advantage: it’s fast. which means there is plenty of publicly available data out there to use in training AI models. When combined with human expertise, AI could help speed up a painstaking process. What’s next isn’t clear, but the prospects are exciting. “Part of it is also just exploring the space and figuring out what potential services or business models might be,” Fuhrer says. Melissa Heikkilä is a senior reporter at MIT Technology Review, covering artificial intelligence and how it is changing our society. 7/31/23 10:58 AM 12 The Download What’s the frequency? Visualizing the beautiful complexity of the United States Frequency Allocation Chart By Jon Keegan Somewhere above you right now, a plane is broadcasting its coordinates on 1090 megahertz. A satellite high above Earth is transmitting weather maps on 1694.1 MHz. On top of all that, every single phone and Wi-Fi router near you blasts internet traffic through the air over radio waves. A carefully regulated radio spectrum is what makes it possible for these signals to get to the right place intact. The Federal Communication Commission and the National Telecommunications and Information Administration share the task of managing radio frequencies for US airwaves. The NTIA manages all federal radio uses (including military use), while the FCC manages everything else. It is an incredibly complex system, SO23-front_thedownload.indd 12 and to help with the job of explaining the importance of managing this invisible natural resource, the NTIA publishes the United States Frequency Allocation Chart (which you can order as a wall chart for $6). The US government lays claim to a large chunk of spectrum for military use, communications, and transportation. FM radio operates between 88 and 108.0 MHz, and AM radio operates between 540 and 1700 kilohertz. Using licenses, amateur radio operators are granted slices where they can communicate safely, as are businesses and other institutions. Civil aviation, maritime navigation, satellite communications, radio astronomy, cellular voice, and data all lay claim to colorful plots on this chart. The chart uses 33 color-coded categories to visualize the information in a crazy quilt of blocks (some wide, some narrow), spread from 9 kHz (very low frequency) all the way to 300 GHz (extremely high frequency). It does suffer from scale distortions, not unlike a map of Earth. Eric Rosenberg, a telecommunications specialist at NTIA, says a lot of the choices about what service goes where come down to physics and the environment where the service will be used: “You can’t just pick up a block and say, okay, we’re gonna move these radars over here.” The chart is always extremely popular, Rosenberg says; fans include lawmakers in Congress. Last updated in 2016, it is due for another revision. “We’re getting to the point where we really feel that we need to redo it,” he says. “Again, it’s a very large project.” A version of this story appeared on Beautiful Public Data (beautifulpublicdata.com), a newsletter by Jon Keegan (KE3GAN). 7/31/23 10:58 AM The Download 13 Above: A detail of the United States Frequency Allocation Chart. SOURCE: NTIA Right: The complete frequency spectrum. SO23-front_thedownload.indd 13 7/31/23 10:58 AM The Download Extracting climate records from Antarctic ice cores Scientists now have the technology to unlock 20,000 years of ancient climate history compressed in a meter of ice. By Christian Elliott Moving quickly and carefully in two layers of gloves, Florian Krauss sets a cube of ice into a gold-plated cylinder that glows red in the light of the aiming laser. He steps back to admire the machine, covered with wires and gauges, that turns polar ice into climate data. If this were a real slice of precious million-year-old ice from Antarctica and not just a test cube, he’d next seal the extraction vessel under a vacuum and power on the 150-megawatt main laser, slowly causing the entire ice sample to sublimate directly into gas. For Krauss, a PhD student at the University of Bern in Switzerland, this would unlock its secrets, exposing the concentrations of greenhouse gases like carbon dioxide trapped within. To better understand the role atmospheric carbon dioxide plays in Earth’s climate cycles, scientists have long turned to ice cores drilled in Antarctica, where snow layers accumulate and compact over SO23-front_thedownload.indd 14 hundreds of thousands of years, trapping samples of ancient air in a lattice of bubbles that serve as tiny time capsules. By analyzing those bubbles and the ice’s other contents, like dust and water isotopes, scientists can connect greenhouse-gas concentrations with temperatures going back 800,000 years. Europe’s Beyond EPICA (European Project for Ice Coring in Antarctica) initiative, now in its third year, hopes to eventually retrieve the oldest core yet, dating back 1.5 million years. This would extend the climate record all the way back to the Mid-Pleistocene Transition, a mysterious period that marked a major change in the frequency of Earth’s climatic oscillations—cycles of repeating glacial and warm periods. Successfully drilling a core that old—a years-long endeavor—might be the easy part. Next, scientists must painstakingly free the trapped air from that ice. Krauss and his colleagues are developing an innovative new way to do that. “We’re not interested in the ice itself— we’re just interested in the air samples included, so we needed to find a new way to extract the air from the ice,” he says. Melting isn’t an option because carbon dioxide easily dissolves into water. Traditionally, scientists have used mechanical extraction methods, grinding up samples of individual layers of ice to free the air. But grinding wouldn’t be effective for the Beyond EPICA ice in the university’s storage freezer, which is kept at 50 °C below zero. The oldest ice at the very bottom of the core will be so compressed, and the individual annual layers so thin, that bubbles won’t be visible—they’ll have been pressed into the lattice of ice crystals, forming a new phase called clathrate. “At the very bottom, we expect 20,000 years of climate history compressed in only one meter of ice,” says Hubertus Fischer, head of the past climate and ice core science group at Bern. That’s a hundredth the thickness of any existing ice core record. The new method Krauss and Fischer are developing is called deepSLice. (A pizza menu is taped to the side of the device right under the laser warning labels, a gift from a pizzeria in Australia with the same name.) DeepSLice has two parts. The Laser-Induced Sublimation Extraction Device, or LISE, fills half a room in the team’s lab space. LISE aims a near-infrared COURTESY PHOTOS 14 An ice core sample (above); Fischer (right) and Krauss with their LISE apparatus. 7/31/23 10:58 AM COURTESY OF THE PUBLISHERS The Download laser continuously at a 10-centimeter slice of ice core so that it turns directly from solid to gas under extremely low pressure and temperature. The sublimated gas then freezes into six metal dip tubes cooled to 15 K (-258 °C), each containing the air from one centimeter of ice core. Finally the samples are loaded into a custommade absorption spectrometer based on quantum cascade laser technology, which shoots photons through the gas sample to measure concentrations of carbon dioxide, methane, and nitrous oxide simultaneously. Another big advantage of this system is that it takes a lot less ice (and work) than the old method of analysis, in which scientists measured methane by melting ice (it doesn’t dissolve into water) and measured carbon dioxide by grinding ice. DeepSLice offers “a unique capability that nobody else has,” says Christo Buizert, an ice core scientist at the University of Oregon and the ice analysis lead for COLDEX (the Center for Oldest Ice Exploration)—the US equivalent of Beyond EPICA, which is currently in a “friendly race” with the Europeans to drill a continuous core down to 1.5-millionyear-old ice. “What they’re trying to do, sublimating ice—people have been trying this for a long time, but it’s one of the most challenging ways to extract gases from ice,” Buizert says. “It’s a very promising way, because you get 100% of the gases out, but it’s very difficult to do. So the fact that they’ve managed to get it working is very impressive.” Krauss and Fischer still have about three years before they get their hands on that section of critical ice. There are still kinks to iron out, like how to recapture the samples from the spectrometer for additional analysis, but they think they’ll be ready when it finally arrives in freezer containers on a ship from Antarctica via Italy. “Our latest results showed us we are on a good track, and actually, we achieved the precision we wanted to,” Krauss says. “So I’m sure it’s going to be ready.” Christian Elliott is a science and environmental reporter based in Chicago. SO23-front_thedownload.indd 15 15 Book reviews Nuts & Bolts: Seven Small Inventions That Changed the World (in a Big Way) By Roma Agrawal (W.W. Norton, 2023) Months spent taking apart ballpoint pens and blenders led Agrawal, an engineer, to explore how seven fundamental inventions led to our most complex feats of engineering. Despite its complexity, she writes, engineering at its most fundamental “is inextricably linked to your everyday life and to humanity.” The Philosopher of Palo Alto: Mark Weiser, Xerox PARC, and the Original Internet of Things By John Tinnell (University of Chicago Press, 2023) “The ‘Smart House’ of 2005 will have computers in every room,” wrote Mark Weiser in 1996, “but what will they do?” The first chief technology officer of Xerox PARC and the so-called father of ubiquitous computing, Weiser (who died at 46 in 1999) was wildly innovative—and prescient. But his vision for the Internet of Things didn’t work out as he’d hoped: the technology meant to connect and lift up humanity instead began to surveil and sell to us. Mobility By Lydia Kiesling (Crooked Media, 2023) An American teenager living in Azerbaijan with her Foreign Service family in the ’90s, Bunny finds herself adrift when she returns to America. She gets a temp job at a Texas oil company—and never leaves. In this novel, Kiesling charts the arc of Bunny’s career (which Bunny always insists is not in oil but at “an energy company”) over two decades, slyly inserting a narrative of our collective apathy toward climate change. The Apple II Age: How the Computer Became Personal By Laine Nooney (University of Chicago Press, 2023) If you want to understand how Apple became an industry behemoth, says Nooney, look no further than the 1977 Apple II. Nooney is keen to critique the lone-genius narrative that characterizes so much of technological advancement, arguing that above all, the story of personal computing in the United States is about the rise of everyday users. 7/31/23 10:58 AM 16 The Download How saving Venice’s salt marshes could keep the city from sinking The Venice lagoon is an ideal test case for new approaches to combating climate change. Venice, Italy, is suffering from a combination of subsidence—the city’s foundations slowly sinking into the mud on which they are built—and rising sea levels. In the worst-case scenario, it could disappear underwater by the year 2100. Alessandro Gasparotto, an environmental engineer, is one of the many people trying to keep that from happening. Standing on a large mudflat in the center of the Venetian lagoon, he pushes a hollow three-foot-high metal cylinder called a piezometer into the thick black mud. This instrument will measure how groundwater moves through the sediment as the lagoon’s tides rise and fall. Knowing what’s happening under the mud is crucial for understanding whether, and how, vegetation can grow and eventually transform this barren landscape of mud into a salt marsh. Gasparotto’s work with salt marshes is part of a project steered by the NGO We Are Here Venice (WAHV) and funded by the EU through the WaterLANDS research program, which is restoring wetlands across Europe. The Venice chapter has been granted €2 million over five years to investigate whether artificial mudflats— the deposits that result when the lagoon is dredged to create shipping channels— can be turned back into the marshes that once thrived in this area and become a functioning part of the lagoon ecosystem again. “The history of the city of Venice has always been intertwined with the history of the lagoon,” explains Andrea D’Alpaos, a geoscientist at the University of Padova. The health of Venice depends on the health of the lagoon system, and vice versa. SO23-front_thedownload.indd 16 This relationship is not only economic—protecting the lagoon ecosystem bolsters fishing yields, for example—but also infrastructural. Salt marshes have a buffering effect on tidal currents, attenuating the force of waves and reducing the water’s erosive effect on Venice’s buildings. But the marshes have been declining for centuries. This is due in part to waterway mismanagement going as far back as the 1500s, when Venetians diverted rivers out of the lagoon, starving it of sediment that would naturally be borne in on their currents. The building of breakwaters at three inlets on the Adriatic Sea and the excavation of an enormous shipping canal in the late 1900s further eroded the marshland. And while the city has been the beneficiary of thousands of euros in restoration and prevention work—most notably the €6.2 billion MOSE (the Italian acronym for “Experimental Electromechanical Module”), a colossal (and extremely effective) system of mobile sea barriers designed to keep the Adriatic’s floodwaters from the city—the marshes have been overlooked. Construction of MOSE began in 2003, but delays, cost overruns and a corruption scandal stalled its completion. It was activated for the first time, successfully preventing a flood, in 2020. Paradoxically, it is the MOSE technology, which protects the city, that is damaging the lagoon ecosystem. “When the MOSE system is raised, it stops storm surges and prevents Venice flooding,” D’Alpaos says. “Storm surges are bad for Venice, but they are good for marshes; 70% of sediment that reaches the marsh is delivered during storm surges.” These excessively high tides, D’Alpaos continues, are happening more often. The problem, he says, is that “if you close the lagoon too often or for too long, you prevent sediment reaching marshes.” In the more than 20 years that he has been studying the lagoon, he says, he’s seen marshes disappearing at an alarming rate: “The marshes are drowning. Two centuries ago, the Venice lagoon had 180 square kilometers [69 square miles] of marshes. Now we only have 43 square kilometers.” One of the sites the We Are Here Venice team is working is on a natural salt marsh, hugged on one side by a kidneyshaped platform of infill dredged from the lagoon. In places where the mud is dry, the ground has separated into patches that conjure small tectonic plates, littered with bone-white crab claws picked clean and dropped by gulls flying overhead. Three orange sticks mark the spot where a fence between the salt marsh and the infill will be removed to allow water exchange and the movement of sediment, making the two ecosystems “speak to one another,” as Jane da Mosto, the executive director and cofounder of WAHV, describes it. COURTESY PHOTO By Catherine Bennett 7/31/23 10:58 AM The Download Jane da Mosto and Alessandro Gasparotto survey Venice’s central lagoon from a restored salt marsh. 17 Jobs of the future: Chief heat officer It’s becoming an essential new role as more cities are beset by extreme heat. COURTESY PHOTO By Allison Arieff Tramping over the island in rubber boots, releasing gobbets of black mud at every step, da Mosto explains that “all of this represents a kind of natural capital.” Not only do the marshes store carbon, but “these environments also support fish habitats and a huge bird population,” she adds. Even the samphire, an edible marshland plant, “could be cultivated like a crop.” Marshes are also more efficient carbon sinks than forests, because marshland plants that store carbon are gradually buried under sediment as the tide washes over them, trapping the carbon for as long as centuries. Da Mosto sees the city as something of a laboratory for environmental solutions with wider applications. “Venice is a mirror on the world,” she says. “If the city remains an example of all the world’s problems, as it is now, then there’s no point trying to keep it alive. But we should be able to show how to turn infills into ecologically productive salt marshes and how to transform an economy based on mass tourism into an economy based on its natural capital.” Catherine Bennett is a freelance journalist based in Paris. SO23-front_thedownload.indd 17 In Miami, extreme heat is a deadly concern. Rising temperatures now kill more people than hurricanes or floods, and do more harm to the region’s economy than rising sea levels. That’s why, in 2021, Florida’s Miami-Dade County hired a chief heat officer, Jane Gilbert—the first position of its kind in the world. Heat has been a silent killer in Miami, says Gilbert: “The number-one cause of weather-related death is from excess heat. It’s been this underrecognized issue that needs to be elevated.” According to the Centers for Disease Control and Prevention, there are an average of 67,512 emergency department visits in the US due to heat each year, and 702 heat-related deaths. A holistic approach: Gilbert works in the county’s Office of Resilience, which has people designated to work on sea-level rise, carbon mitigation, and waste reduction. “Together,” she says, “we make sure we come at it from an integrated perspective.” She acknowledges that some may be skeptical of her role because “if you work and live in air-conditioning and can afford it, you can manage heat, [and] you don’t need me.” Inform, prepare, protect: Gilbert’s focus is on those least able to protect themselves and their families against high heat—poorer communities and Black and Hispanic people tend to bear the brunt. Her collaborative efforts to keep homes, facilities, and neighborhoods affordably cool include everything from creating programs that protect outdoor workers to planting trees that help mitigate heat-island effects. Career path: Gilbert majored in environmental science at Barnard College in New York City and went on to get a master’s in public administration at Harvard’s Kennedy School of Government, focusing on urban community development. The job of chief heat officer didn’t exist back then, she says, but if it had, “I would have been really interested.” Some of the issues may have shifted, she explains, “but when I studied climate change in the mid-’80s, it was accepted science.” 7/31/23 10:58 AM 18 Explained How french fries, trash, and sunlight could power your future flights. By Casey Crownhart Illustration by Marcin Wolski Everything you need to know about the wild world of alternative jet fuels Aviation accounts for about 2% of global carbon dioxide emissions, and considering the effects of other polluting gases, the industry is responsible for about 3% of all human-caused global warming. One way the aviation industry hopes to cut down on its climate impacts is by using new fuels. These alternatives, often called sustainable aviation fuels (SAFs), could be the key to helping this sector reach net-zero carbon dioxide emissions by 2050. The actual climate impact of alternative fuels will depend on a lot of factors, however. Here’s everything you need to know about the future of jet fuel and the climate. What are SAFs? Planes today mostly burn kerosene—a fossil fuel with a mix of carbon-containing molecules. Alternative fuels have the same basic chemical makeup as traditional jet fuel, but they are derived from renewable sources. Alternative fuels fall into two main categories: biofuels and synthetic electrofuels. Biofuels come from a range of biological sources; some are derived from waste like used cooking oil, agricultural residues, or household trash, while others are made from crops like corn and palm trees. Making fuel from biological sources requires chopping up the complicated chemical structures that plants make to store energy. Fats and carbohydrates can be broken apart and purified to make the simple chains of carbon-rich molecules that are jet fuel’s primary ingredient. Electrofuels (also called e-fuels), on the other hand, don’t start with plants. Instead, they start with two main building blocks: hydrogen and carbon dioxide, which are SO23-front_explained.indd 18 combined and transformed in chemical reactions powered by electricity. Making e-fuels is expensive today, because the process is inefficient and isn’t done widely at commercial scale. But experts say that to reach its 2050 target, aviation will need to rely on them, because they’re the most effective way to cut carbon dioxide emissions, and they won’t be limited by supply or collection logistics like fuels made from plants or waste. So how do SAFs help climate progress? Like conventional jet fuel, alternative fuels produce carbon dioxide when they’re burned for energy in planes. Unlike regular airplanes, those that run on SAFs can, depending on how the fuels are made, offset their carbon dioxide emissions. In an ideal world, a fuel’s production process would remove enough carbon from the atmosphere to cancel out the carbon dioxide emissions when the fuel is burned. However, that’s often far from the reality. Alternative fuels fall on a spectrum in terms of how much they reduce carbon dioxide emissions. On one end, synthetic fuels that are made with carbon collected via direct air capture and whose production is powered entirely by renewable electricity will reduce emissions by nearly 100% compared with fossil fuels. On the other end of the spectrum, some crop-based biofuels can produce more carbon dioxide emissions overall than fossil fuels. That’s frequently the case for biofuels made from palm oil, since growing that crop can decimate rainforests. 7/28/23 10:28 AM Explained Today, most commercially available alternative jet fuels are made from fats, oils, and greases. If they’re derived from waste sources like these, such fuels reduce carbon dioxide emissions by roughly 70% to 80% compared with fossil fuels. It’s worth noting that while SAFs can approach net-zero carbon dioxide emissions, burning the fuels still produces other pollution and contributes to contrails, which can trap heat in the atmosphere. What’s next for SAFs? Alternative fuels are attractive to the aviation industry because they’re a drop-in SO23-front_explained.indd 19 solution, requiring little adjustment of aircraft and airport infrastructure. Over the past year, several test flights powered by 100% SAFs have taken off. However, alternative fuels made up less than 0.2% of the global jet fuel supply in 2022. One of the main challenges to getting SAFs into the skies is expanding the supply. The world doesn’t eat enough french fries for used cooking oils to meet global demand for jet fuel. Recent policy moves in both the United States and the European Union are aimed at boosting the market for alternative fuels. RefuelEU Aviation, a deal finalized 19 in April, requires that fuel supply at EU airports include 2% SAFs by 2025 and 70% by 2050. The US recently passed new tax credits for alternative fuels, aimed at helping expensive options reach price parity with fossil fuels. Ultimately, alternative fuels present one potential pathway to cutting the climate impacts of aviation. But the details will matter profoundly: some fuels could be part of the solution, while others might end up being part of the problem. Casey Crownhart is a climate reporter at MIT Technology Review. 7/28/23 10:28 AM 20 Profile Valley of the misfit tech workers F or Silicon Valley venture capitalists and founders, any inconvenience big or small is a problem to be solved—even death itself. And a new genre of products and services known as “death tech,” intended to help the bereaved and comfort the suffering, shows that the tech industry will try to address literally anything with an app. Xiaowei Wang, a technologist, author, and organizer based in Oakland, California, finds that disturbing. “It’s so gross to view people like that—to see situations and natural facts of life like dying as problems,” Wang said during lunch and beers on the back patio of an Oakland brewery in late March. To research a forthcoming book on the use of tech in end-of-life care, Wang has trained as a “death doula” and will soon start working at a hospice. This approach to exploring technology, grounded in its personal and political implications, exemplifies a wider vision for fellow tech workers and the industry at large—a desire that it grant more power and agency to those with diverse backgrounds, become more equitable instead of extractive, and aim to reduce structural inequalities rather than seeking to enrich shareholders. To realize this vision, Wang has launched a collaborative learning project called Collective Action School in which tech workers can begin to confront their own impact on the world. The hope is to promote more labor organizing within the industry and empower workers who may feel intimidated to challenge gigantic corporations. SO23-front_profile.indd 20 Wang came to prominence as an editor at Logic magazine, an independent publication created in 2016 amid early Trump-era anxiety and concerns about the growing powers of technology. Dismissing utopian narratives of progress for prescient analysis of tech’s true role in widening inequity and concentrating political power, the founders—who also included Ben Tarnoff, Jim Fingal, Christa Hartsock, and Moira Weigel—vowed to stop having “stupid conversations about important things.” (In January, it was relaunched as “the first Black, Asian, and Queer tech magazine,” with Wang and J. Khadijah Abdurahman as co-editors.) Collective Action School, initially known as Logic School, is an outgrowth of the magazine. It’s emerged at a time when scandals and layoffs in the tech industry, combined with crypto’s troubles and new concerns about bias in AI, have made Big Tech’s failings all the more visible. In courses offered via Zoom, Wang and other instructors guide roughly two dozen tech workers, coders, and project managers through texts on labor organizing, intersectional feminist theory, and the political and economic implications of Big Tech. Its second cohort has now completed the program At our lunch, Wang was joined by three former students who helped run that last session: Derrick Carr, a senior software engineer; Emily Chao, a former trust and safety engineer at Twitter; and Yindi Pei, a UX designer. All shared a desire to create something that could lead to more concrete change than existing corporate employee resource groups, which they say often seem constrained and limited. And Xiaowei Wang and Collective Action School seek to remedy the moral blindness of Big Tech. By Patrick Sisson Portrait by Christie Hemm Klok while Big Tech may obsess over charismatic founders, Collective Action School runs in a collective fashion. “I enjoy operating under the radar,” Wang said. W ang, who uses the pronoun “they,” moved from China to Somerville, Massachusetts, in 1990, at age four. Drawn to science and technology at a young age, they made friends in early online chat rooms and built rockets and studied oceanography at science camps. They also started questioning social norms early on; their mom tells of getting a call from the middle school principal, explaining that Wang had started a petition for a gender-inclusive class dress code. Years later, they enrolled at Harvard to study design and landscape architecture—at one point lofting a kite over the skies in Beijing to track pollution levels. A few years after graduating in 2008, Wang moved to the Bay Area. They worked at the nonprofit Meedan Labs, which develops open-source tools for journalists, and the mapping software company Mapbox, a rapidly scaling “rocket ship” where an employee—sometimes Wang—had to be on call, often overnight, to patch any broken code. Unsatisfied, Wang left in 2017 to focus on writing, speaking, and research, earning a PhD in geography at Berkeley. “The person who did my [Mapbox] exit interview told me, ‘You have this problem where you see injustice and you can’t stand it,’” Wang says. “She told me, ‘Sometimes you need to put that to bed if you want to stay in this industry.’ I can’t.” Many in tech, Wang says, have a fundamental belief in constant improvement through corporate innovation; for these 7/27/23 10:51 AM A Buddhist teacher told Wang that we’re all “looking at the sky through a straw,” limited to our own small portholes of perception. This insight guides their approach to research and writing. SO23-front_profile.indd 21 7/27/23 10:51 AM 22 Profile people, technology means “you push a button and something in your life is solved.” But Wang, who practices Buddhism and reads tarot cards, sees things differently, believing that life is all about natural cycles humans can’t control and should accept with humility. For Wang, tech can be rural communities hacking open-source software, or simply something that brings pure joy. At Logic, Wang penned a popular column, Letter from Shenzhen, which included scenes from their family’s hometown of Guangzhou, China, and the explosion of innovation in the country. It led to a book titled Blockchain Chicken Farm: And Other Stories of Tech in China’s Countryside, a striking exploration of technology’s impact on rural China. During the book editing process, Wang went on a Buddhist retreat, where a teacher remarked that we’re all “looking at the sky through a straw,” limited to our own small portholes of perception. This insight, says Wang, helped frame the final draft. But it also became a metaphor for an entire approach to research and writing on technology: focused, careful consideration of many viewpoints, and the capacity to imagine something better. C ollective Action School, funded in part by the Omidyar Network and a grant from the arts and coding nonprofit Processing Foundation, came together in 2020 as tech worker activism was on the rise. Kickstarter employees’ union drive in 2020 was followed by efforts at Alphabet, Amazon, and Apple, as well as industry-wide campaigns such as Collective Action in Tech (led in part by former Logic editor Tarnoff) and the Tech Workers Coalition. But because Wang avoids the spotlight and believes that only strong communities can remedy the tech industry’s ills, the school is organized in a more experimental way. Each cohort begins with a “week zero” meeting to get acquainted as a group. Then, for 13 weeks, participants attend sessions covering labor movements, the political economy of innovation, and the impact SO23-front_profile.indd 22 Collective Action School offers an antithesis to the “golden ticket” mentality of tech work, with an approach that’s more focused on collective action and culture. of technology on marginalized groups. The funding covers all tuition costs for all students. As Pei, one of the co-organizers, puts it, the school offers an antithesis to the “golden ticket” mentality of tech work, with an approach that’s more focused on collective action and culture. Each week, participants read from a lengthy syllabus and welcome a guest speaker. Past guests include Clarissa Redwine from the Kickstarter union’s oral history project, former Google employees Alex Hanna and Timnit Gebru of the Distributed AI Research Institute, and Erin McElroy, cofounder of the Anti-Eviction Mapping Project. Then they work on a final project; one of the first was Looking Glass, which used augmented reality to highlight the lost Black history of Pittsburgh. For developing it, creator Adrian Jones was named the school’s “community technologist,” a role that comes with a one-year grant to expand the idea. Chao, who formerly worked for Twitter, released a zine about trust and safety issues, and Pei has been working on an affordable housing website for San Francisco. The organizers see Collective Action School as a community-building project, and open-source syllabus, that can grow with each new cohort. Eventually, the aim is to expand the reach of the school with chapters based in other areas, adding in-person meetings and creating a larger network of workers sharing similar values and aims. That strategy fills a need within larger tech and labor organizing, says Gershom Bazerman, who volunteers with the Tech Workers Coalition and Emergency Workplace Organizing Committee. Tech workers have long been told they’re unique, but recent political fights between workers and leadership—with employees pushing back against contributing to projects used by the US military or immigration enforcement—have set off a wave of ground-up organizing informed by social concerns. Groups like Collective Action School can be a “bridge” between workers seeking such change. While the readings and interactions aren’t creating a utopia, they are creating a space for students to learn, meet, and commit to more change. Wang hopes they find solidarity and, ideally, bring these ideas and experience back to their companies and coworkers (or find the resources and momentum to move to a job or field more aligned with their values). Some in this year’s cohort live and work in the Global South and have faced layoffs, so classmates created a cost-of-living support fund to help. Carr has called the experience an “antidote to a specific accumulated toxin” that comes from working in Big Tech. That may be true, but Collective Action School, along with other recent organizing efforts, also sets out to redefine the experience of working within the industry. “We’re not saying we’re making the perfect safe learning space,” says Wang. “We had a container in which we could have fun, learn from each other, and then grow. I think that’s really rare and special. It’s like committing to each other.” Patrick Sisson, a Chicago expat living in Los Angeles, covers technology and urbanism. 7/28/23 12:49 PM Expand your knowledge beyond the classroom. Invest in your future with a student subscription and save 50% on year-long access to MIT Technology Review’s trusted reporting, in-depth stories, and expert insights. Scan this code to access your 50% student savings or learn more at TechnologyReview.com/StudentOffer Untitled-3 1 8/3/23 10:57 24 SO23-feature_Hamzelou.therapies.indd 24 8/2/23 6:14 PM 25 By Illustration Jessica Hamzelou Selman Design COVER STORY Desperate people will often want to try experimental, unproven treatments. How can we ensure they’re not exploited or put at risk? SO23-feature_Hamzelou.therapies.indd 25 THE RIGHT TO TRY 8/2/23 6:14 PM 26 Max was only a toddler when his parents noticed there was “some- people to access treatments that might not help them—and thing different” about the way he moved. He was slower than could harm them. Anecdotes appear to be overpowering evidence in decisions on drug approval. As a result, we’re other kids his age, and he struggled to jump. He couldn’t run. ending up with some drugs that don’t work. We urgently need to question how these decisions are Blood tests suggested he might have a genetic disease— made. Who should have access to experimental theraone that affected a key muscle protein. Max’s dad, Tao Wang, pies? And who should get to decide? Such questions are a researcher for a climate philanthropy organization, says especially pressing considering how quickly biotechnology is advancing. Recent years have seen an explosion in he and his wife were initially in denial. It took them a few what scientists call “ultra-novel” therapies, many of which months to take Max for the genetic test that confirmed their involve gene editing. We’re not just improving on existing classes of treatments—we’re creating entirely new ones. fears: he had Duchenne muscular dystrophy. Managing access to them will be tricky. Just last year, a woman received a CRISPR treatment Duchenne is a rare disease that tends to affect young boys. It’s progressive—those affected generally lose muscle designed to lower her levels of cholesterol—a therapy that function as they get older. There is no cure. Many people directly edited her genetic code. Also last year, a genetically with the disorder require wheelchairs by the time they modified pig’s heart was transplanted into a man with severe reach their 20s. Most do not survive beyond their 30s. heart disease. Debates have raged over whether he was the Max’s diagnosis hit Wang and his wife “like a tornado,” right candidate for the surgery, since he ultimately died. he says. But eventually one of his doctors menFor many, especially those with severe diseases, tioned a clinical trial that he was eligible for. The trying an experimental treatment may be better Between 2009 and 2022, trial was for an experimental gene therapy designed than nothing. That’s the case for some people to replace the missing muscle protein with a shortwith Duchenne, says Hawken Miller, a 26-yearened, engineered version that might help slow his old with the condition. “It’s a fatal disease,” he decline or even reverse it. Enrolling Max in the says. “Some people would rather do something CANCER DRUGS trial was a no-brainer for Wang. “We were willing than sit around and wait for it to take their lives.” received accelerated to try anything that could change the course [of the approval to treat 66 disease] and give us some hope,” he says. Expanding access conditions—and 15 of those approvals have That was more than two years ago. Today, Max There’s a difficult balance to be reached between since been withdrawn. is an active eight-year-old, says Wang. He runs, protecting people from the unknown effects of a jumps, climbs stairs without difficulty, and even new treatment and enabling access to something enjoys hiking. “He’s a totally different kid,” says Wang. potentially life-saving. Trying an experimental drug could The gene therapy he received was recently considcure a person’s disease. It could also end up making no ered for accelerated approval by the US Food and Drug difference, or even doing harm. And if companies strugAdministration. Such approvals, reserved for therapies gle to get funding following a bad outcome, it could delay targeting serious conditions that lack existing treatments, progress in an entire research field—perhaps slowing require less clinical trial data than standard approvals. future drug approvals. While the process can work well, it doesn’t always. In the US, most experimental treatments are accessed And in this case, the data is not particularly compelling. through the FDA. Starting in the 1960s and ’70s, drug The drug failed a randomized clinical trial—it was found manufacturers had to prove to the agency that their prodto be no better than a placebo. ucts actually worked, and that the benefits of taking them Still, many affected by Duchenne are clamoring for would outweigh any risks. “That really closed the door on patients’ being able to access drugs on a speculative access to the treatment. At an FDA advisory committee meeting in May set up to evaluate its merits, multiple basis,” says Christopher Robertson, a specialist in health parents of children with Duchenne pleaded with the law at Boston University. It makes sense to set a high bar of evidence for new organization to approve the drug immediately—months before the results of another clinical trial were due. On medicines. But the way you weigh risks and benefits can June 22, the FDA granted conditional approval for the change when you receive a devastating diagnosis. And it drug for four- and five-year-old boys. wasn’t long before people with terminal illnesses started This drug isn’t the only one to have been approved on asking for access to unapproved, experimental drugs. weak evidence. There has been a trend toward lowering In 1979, a group of people with terminal cancer and the bar for new medicines, and it is becoming easier for their spouses brought a legal case against the government 48 SO23-feature_Hamzelou.therapies.indd 26 8/2/23 6:14 PM 27 to allow them to access an experimental treatment. While a district court ruled that one of the plaintiffs should be allowed to buy the drug, it concluded that whether a person’s disease was curable or not was beside the point— everyone should still be protected from ineffective drugs. The decision was eventually backed by the Supreme Court. “Even for terminally ill patients, there’s still a concept of safety and efficacy under the statute,” says Robertson. Today, there are lots of ways people might access experimental drugs on an individual basis. Perhaps the most obvious way is by taking part in a clinical trial. Early-stage trials typically offer low doses to healthy volunteers to make sure new drugs are safe before they are offered to people with the condition the drugs are ultimately meant to treat. Some trials are “open label,” where everyone knows who is getting what. The gold standard is trials that are randomized, placebo controlled, and blinded: some volunteers get the drug, some get the placebo, and no one—not even the doctors administering the drugs— knows who is getting what until after the results have been collected. These are the kinds of studies you need to do to tell if a drug is really going to help people. But clinical trials aren’t an option for everyone who might want to try an unproven treatment. Trials tend to have strict criteria about who is eligible depending on their age and health status, for example. Geography and timing matter, too—a person who wants to try a certain drug might live too far from where the trial is being conducted, or might have missed the enrollment window. Instead, such people can apply to the FDA under the organization’s expanded access program, also known as “compassionate use.” The FDA approves almost all such requests. It then comes down to the drug manufacturer to decide whether to sell the person the drug at cost (it is not allowed to make a profit), offer it for free, or deny the request altogether. Another option is to make a request under the Right to Try Act. The law, passed in 2018, establishes a new route for people with life-threatening conditions to access experimental drugs—one that bypasses the FDA. Its introduction was viewed by many as a political stunt, given that the FDA has rarely been the barrier to getting hold of such medicines. Under Right to Try, companies still have the choice of whether or not to provide the drug to a patient. When a patient is denied access through one of these pathways, it can make headlines. “It’s almost always the same story,” says Alison Bateman-House, an ethicist who researches access to investigational medical products at New York University’s Grossman School of Medicine. In this story, someone is fighting for access to a drug and being denied it by “cold and heartless” pharma or the FDA, she says. The story is always about “patients valiantly struggling for something that would undoubtedly help them if they could just get to it.” But in reality, things aren’t quite so simple. When companies decide not to offer someone a drug, you can’t really blame them for making that decision, says BatemanHouse. After all, the people making such requests are usually incredibly ill. If someone were to die after taking that drug, not only would it look bad, but it could also put off investors from funding further development. “If you have a case in the media where somebody gets compassionate use and then something bad happens to them, investors run away,” says Bateman-House. “It’s a business risk.” FDA approval of a drug means it can be sold and prescribed—crucially, it’s no longer experimental. Which is why many see approval as the best way to get hold of a promising new treatment. “If ... somebody gets compassionate use and then something bad happens to them, investors run away. It’s a business risk.” SO23-feature_Hamzelou.therapies.indd 27 As part of a standard approval process, which should take 10 months or less, the FDA will ask to see clinical trial evidence that the drug is both safe and effective. Collecting this kind of evidence can be a long and expensive process. But there are shortcuts for desperate situations, such as the outbreak of covid-19 or rare and fatal diseases—and for serious diseases with few treatment options, like Duchenne. Anecdotes vs. evidence Max accessed his drug through a clinical trial. The treatment, then called SRP-9001, was developed by the pharmaceutical company Sarepta and is designed to replace dystrophin, the protein missing in children with Duchenne muscular dystrophy. The protein is thought to protect muscle cells from damage when the muscles contract. Without it, muscles become damaged and start to degenerate. The dystrophin protein has a huge genetic sequence— it’s too long for the entire thing to fit into a virus, the usual means of delivering new genetic material into a person’s body. So the team at Sarepta designed a shorter version, which they call micro-dystrophin. The code for the protein 8/2/23 6:14 PM 28 is delivered by means of a single intravenous infusion. The company’s initial goal was to develop it as a therapy for children between four and seven with a diagnosis of Duchenne. And it had a way to potentially fast-track the process. Usually, before a drug can be approved, it will go through several clinical trials. But accelerated approval offers a shortcut for companies that can show that their drug is desperately needed, safe, and supported by compelling preliminary evidence. For this kind of approval, drug companies don’t need to show that a treatment has improved anyone’s health—they just need to show improvement in some biomarker related to the disease (in Sarepta’s case, the levels of the micro-dystrophin protein in people’s blood). There’s an important proviso: the company must promise to continue studying the drug, and to provide “confirmatory trial evidence.” This process can work well. But in recent years, it has been a “disaster,” says Diana Zuckerman, president of the National Center for Health Research, a nonprofit that assesses research on health issues. Zuckerman believes the bar of evidence for accelerated approval has been dropping. Many drugs approved via this process are later found ineffective. Some have even been shown to leave people worse off. For example, between 2009 and 2022, 48 cancer drugs received accelerated approval to treat 66 conditions—and 15 of those approvals have since been withdrawn. Melfulfen was one of these. The drug was granted accelerated approval for multiple myeloma in February 2021. Just five months later, the FDA issued an alert following the release of trial results suggesting that people taking the drug had a higher risk of death. In October 2021, the company that made the drug announced it was to be taken off the market. There are other examples. Take Makena, a treatment meant to reduce the risk of preterm birth. The drug was granted accelerated approval in 2011 on the basis of results from a small trial. Larger, later studies suggested it didn’t work after all. Earlier this year, the FDA withdrew approval for the drug. But it had already been prescribed to hundreds of thousands of people—nearly 310,000 women were given the drug between 2011 and 2020 alone. And then there’s Aduhelm. The drug was developed as a treatment for Alzheimer’s disease. When trial data was presented to an FDA advisory committee, 10 of 11 panel SO23-feature_Hamzelou.therapies.indd 28 However, the drug being studied, Sarepta's SRP-9001,failed to perform better than placebo across the whole group of boys in the trial. members voted against approval. The 11th was uncertain. There was no convincing evidence that the drug slowed cognitive decline, the majority of the members found. “There was not any real evidence that this drug was going to help patients,” says Zuckerman. Despite that, the FDA gave Aduhelm accelerated approval in 2021. The drug went on the market at a price of $56,000 a year. Three of the committee members resigned in response to the FDA’s approval. And in April 2022, the Centers for Medicare & Medicaid Services announced that Medicare would only cover treatment that was administered as part of a clinical trial. The case demonstrates that accelerated approval is no guarantee a drug will become easier to access. The other important issue is cost. Before a drug is approved, people might be able to get it through expanded access—usually for free. But once the drug is approved, many people who want it will have to pay. And new treatments—especially gene therapies—don’t tend to be cheap. We’re talking hundreds of thousands, or even millions, of dollars. “No patient or families should have to pay for a drug that’s not proven to work,” says Zuckerman. What about SRP-9001? On May 12, the FDA held an advisory committee meeting to assess whether the data PHOTO COURTESY OF TAO WANG More than two years after participating in a clinical trial testing a treatment for Duchenne muscular dystrophy, Max Wang is an active eightyear-old. 8/3/23 8:40 AM 29 PHOTO COURTESY OF ROBERTS FAMILY Will Roberts (left), who is now 10 years old, has been taking Sarepta’s Amondys 45 to treat his Duchenne muscular dystrophy since he was a year old, but he has seen little improvement. His parents, Ryan and Keyan, were hoping that SRP-9001 would be approved by the FDA. supported accelerated approval. During the nine-hour virtual meeting, scientists, doctors, statisticians, ethicists, and patient advocates presented the data collected so far, and shared their opinions. Sarepta had results from three clinical trials of the drug in boys with Duchenne. Only one of the three—involving 41 volunteers aged four to seven—was randomized, blinded, and placebo controlled. Scientists will tell you that’s the only study you can draw conclusions from. And unfortunately, that trial did not go particularly well—by the end of 48 weeks, the children who got the drug were not doing any better than those who got a placebo. But videos presented by parents whose children had taken the drug told a different story. Take the footage shared by Brent Furbee. In a video clip taken before he got the gene therapy, Furbee’s son Emerson is obviously struggling to get up the stairs. He slowly swings one leg around while clinging to the banister, before dragging his other leg up behind him. A second video, taken after the treatment, shows him taking the stairs one foot at a time, with the speed you’d expect of a healthy four-year-old. In a third, he is happily pedaling away on his tricycle. Furbee told the committee SO23-feature_Hamzelou.therapies.indd 29 that Emerson, now six, could run faster, get up more quickly, and perform better on tests of strength and agility. “Emerson continues to get stronger,” he said. It was one of many powerful, moving testimonies—and these stories appear to have influenced the FDA’s voting committee, despite many concerns raised about the drug. The idea of providing the genetic code for the body to make a shortened version of dystrophin is based on evidence that people who have similarly short proteins have a much milder form of muscular dystrophy than those whose bodies produce little to no dystrophin. But it’s uncertain whether Sarepta’s protein, with its missing regions, will function in the same way. Louise Rodino-Klapac, executive vice president, chief scientific officer, and head of R&D at Sarepta, defends the drug: “The totality of the evidence is what gives us great confidence in the therapy.” She has an explanation for why the placebo-controlled trial didn’t show a benefit overall. The groups of six- to seven-year-olds receiving the drug and the placebo were poorly matched “at baseline,” she says. She also says that the researchers saw a statistically significant result when they focused only on the four- and five-year-olds studied. But the difference is not statistically significant for the results the trial was designed to collect. And there are some safety concerns. While most of the boys developed only “mild” side effects, like vomiting, nausea, and fever, a few experienced more serious, although temporary, problems. There were a total of nine serious complications among the 85 volunteers. One boy had heart inflammation. Another developed an immune disease that damages muscle fibers. On top of all that, as things currently stand, receiving one gene therapy limits future gene therapy options. That’s because the virus used to deliver the therapy causes the body to mount an immune response. Many gene therapies rely on a type called adeno-associated virus, or AAV. If a more effective gene therapy that uses the same virus comes along in the coming years, those who have taken this drug won’t be able to take the newer treatment. Despite all this, the committee voted 8–6 in favor of granting the drug an accelerated approval. Many committee members highlighted the impact of the stories and videos shared by parents like Brent Furbee. “Now, I don’t know whether those boys got placebo or whether they got the drug, but I suspect that they got 8/2/23 6:14 PM 30 the drug,” a neurologist named Anthony Amato told the audience. “Those videos, anecdotal as they are … are substantial evidence of effectiveness,” said committee member Donald B. Kohn, a stem-cell biologist. The drugs don’t work? was running around in their backyard, but this year he needs a power chair to get around at school. “We definitely didn’t see any gains in ability, and it’s hard to tell if it made his decline … a little less steep,” Roberts says. The treatment comes with risks, too. The Amondys 45 website warns that 20% of people who get the drug experience adverse reactions, and that “potentially fatal” kidney damage has been seen in people treated with a similar drug. Roberts says she is aware of the risks that come with taking drugs like Amondys. But she and her husband, Ryan, an IT manager, were still hoping that SRP-9001 would be approved by the FDA. For the Robertses and parents like them, part of the desire is based on the hope, no matter how slim, that their child might benefit. “We really feel strongly that we’re in a position now where we’re seeing [Will’s] mobility decline, and we’re nervous that … he might not qualify to take it by the time it’s made available,” she said in a video call, a couple of weeks after the advisory committee meeting. Powerful as they are, individual experiences are just that. “If you look at the evidentiary hierarchy, anecdote is considered the lowest level of evidence,” says Bateman-House. “It’s certainly nowhere near clinical-trial-level evidence.” This is not the way we should be approving drugs, says Zuckerman. And it’s not the first time Sarepta has had a drug approved on the basis of weak evidence, either. The company has already received FDA approval to sell three other drugs for Duchenne, all of them designed to skip over faulty exons—bits of DNA that code for a protein. Such drugs should allow cells to make a longer form of a protein that more closely resembles dystrophin. The first of these “exon-skipping” drugs, Exondys 51, was granted accelerated approval in Selling hope 2016—despite the fact that the clinical trial was not On June 22, just over a month after the committee SRP-9001, placebo controlled and included only 12 boys. “I’ve meeting, the FDA approved SRP-9001, now called now called Elevidys, never seen anything like it,” says Zuckerman. She Elevidys. It will cost $3.2 million for the one-off will cost points out that the study was far too small to be treatment, before any potential discounts. For the able to prove the drug worked. In her view, 2016 time being, the approval is restricted to four- and was “a turning point” for FDA approvals based on five-year-olds. It was granted with a reminder to MILLION the company to complete the ongoing trials and low-quality evidence—“It was so extreme,” she says. for a one-off treatment. report back on the results. Since then, three other exon-skipping drugs have received accelerated approval for Sarepta maintains that there is sufficient eviDuchenne—two of them from Sarepta. A Sarepta dence to support the drug’s approval. But this drug spokesperson said a company-funded analysis showed that and others have been made available—at eye-wateringly people with Duchenne who received Exondys 51 remained high prices—without the strong evidence we’d normally ambulatory longer and lived longer by 5.4 years—“data expect for new medicines. Is it ever ethical to sell a drug we would not have without that initial approval.” when we don’t fully know whether it will work? But for many in the scientific community, that data still I put this question to Debra Miller, mother of Hawken needs to be confirmed. “The clinical benefit still has not Miller and founder of CureDuchenne. Hawken was been confirmed for any of the four,” Mike Singer, a clinical diagnosed when he was five years old. “The doctor that reviewer in the FDA’s Office of Therapeutic Products, told diagnosed him basically told us that he was going to stop the advisory committee in May. walking around 10 years old, and he would not live past “All of them are wanted by the families, but none of 18,” she says. “‘There’s no treatment. There’s no cure. them have ever been proven to work,” says Zuckerman. There’s nothing you can do. Go home and love your child.’” Will Roberts is one of the boys taking an exon-skipping She set up CureDuchenne in response. The organization is dedicated to funding research into potential treatdrug—specifically, Sarepta’s Amondys 45. Now 10, he was diagnosed with Duchenne when he was just one year old. ments and cures, and to supporting people affected by the His treatment involves having a nurse come to his home disease. It provided early financial support to Sarepta but and inject him every five to 10 days. And it’s not cheap. does not have a current financial interest in the company. While his parents have a specialist insurance policy that Hawken, now a content strategist for CureDuchenne, has shields them from the cost, the price of a year’s worth of never been eligible for a clinical trial. treatment is around $750,000. Debra Miller says she’s glad that the exon-skipping Will’s mother, Keyan Roberts, a teacher in Michigan, drugs were approved. From her point of view, it’s about says she can’t tell if the drug is helping him. Last year he more than making a new drug accessible. $3.2 SO23-feature_Hamzelou.therapies.indd 30 8/2/23 6:14 PM 31 “[The approvals] drove innovation and attracted a lot of attention to Duchenne,” she says. Since then, CureDuchenne has funded other companies exploring next-generation exon-skipping drugs that, in early experiments, seem to work better than the first-generation drugs. “You have to get to step one before you can get to step two,” she says. Hawken Miller is waiting for the data from an ongoing phase 3 clinical trial of Elevidys. For the time being, “from a data perspective, it doesn’t look great,” he says. “But at the same time, I hear a lot of anecdotes from parents and patients who say it’s really helping a lot, and I don’t want to discount what they’re seeing.” Results were due in September—just three months after the accelerated approval was granted. It might not seem like much of a wait, but every minute is precious to children with Duchenne. “Time is muscle” was the refrain repeated throughout the advisory committee meeting. “I wish that we had the time and the muscle to wait for things that were more effective,” says Keyan Roberts, Will’s mom. “But one of the problems with this disease is that we might not have the opportunity to wait to take one of those other drugs that might be made available years down the line.” Doctors may end up agreeing that a drug—even one that is unlikely to work— is better than nothing. “In the American psyche, that is the approach that [doctors and] patients are pushed toward,” says Holly Fernandez Lynch, a bioethicist at the University of Pennsylvania. “We have all this language that you’re ‘fighting against the disease,’ and that you should try everything.” “I can’t tell you how many FDA advisory committee meetings I’ve been to where the public-comment patients are saying something like ‘This is giving me hope,’” says Zuckerman. “Sometimes hope helps people do better. It certainly helps them feel better. And we all want hope. But in medicine, isn’t it better to have hope based on evidence rather than hope based on hype?” Their reasoning is that people affected by devastating diseases should be protected from ineffective and possibly harmful treatments—even if they want them. Review boards assess how ethical clinical trials are before signing off on them. Participants can’t be charged for drugs they take in clinical trials. And they are carefully monitored by medical professionals during their participation. That doesn’t mean people who are desperate for treatments are incapable of making good decisions. “They are stuck with bad choices,” says Fernandez Lynch. This is also the case for ultra-novel treatments, says Robertson. At the start of trials, the best candidates for allnew experimental therapies may be those who are closer to death, he says: “It is quite appropriate to select patients who have less to lose, while nonetheless being sure not to exploit people who don’t have any good options.” There’s another advantage to clinical trials. It’s hard to assess the effectiveness of a one-off treatment in any single individual. But clinical trials contribute valuable “We all want hope. But in medicine, isn’t it better to have hope based on evidence rather than hope based on hype?” A desperate decision A drug approved on weak data might offer nothing more than false hope at a high price, Zuckerman says: “It is not fair for patients and their families to [potentially] have to go into bankruptcy for a drug that isn’t even proven to work.” The best way for people to access experimental treatments is still through clinical trials, says Bateman-House. Robertson, the health law expert, agrees, and adds that trials should be “bigger, faster, and more inclusive.” If a drug looks as if it’s working, perhaps companies could allow more volunteers to join the trial, for example. SO23-feature_Hamzelou.therapies.indd 31 data that stands to benefit a patient community. Such data is especially valuable for treatments so new that there are few standards for comparison. Hawken Miller says he would consider taking part in an Elevidys clinical trial. “I’m willing to take on some of that risk for the potential of helping other people,” he says. “I think you’ll find that in [most of the Duchenne] community, everyone’s very willing to participate in clinical trials if it means helping kids get cured faster.” When it comes to assessing the likelihood that Elevidys will work, Will’s dad, Ryan Roberts, says he’s a realist. “We’re really close to approaching the last chance—the last years he’ll be ambulatory,” he says. For him as a dad, he says, the efficacy concerns aren’t relevant. “We will take the treatment because it’s going to be the only chance we have … We are aware that we’re not being denied a treatment that is a cure, or a huge game-changer. But we are willing to take anything we can get in the short window we have closing now.” Jessica Hamzelou is a senior reporter at MIT Technology Review. 8/2/23 6:14 PM 32 ON LY HUMAN JUST BEFORE CHRISTMAS LAST YEAR, A PASTOR PREACHED A GOSPEL OF MORALS OVER MONEY TO SEVERAL HUNDRED MEMBERS OF HIS FLOCK. Wearing a sport coat, angular glasses, and wired earbuds, he spoke animatedly into his laptop from his tiny glass office inside a co-working space, surrounded by six whiteboards filled with his feverish brainstorming. Sharing a scriptural parable familiar to many in his online audience—a group assembled from across 48 countries, many in the Global South—he explained why his congregation was undergoing dramatic growth in an age when the life of the spirit often struggles to compete with cold, hard, capitalism. “People have different sources of motivation [for getting involved in a community],” he sermonized. “It’s not only money. People actually have a deeper purpose in life.” Many of the thousands of people who’d been joining his community were taking the time and energy to do so “because they care about the human condition, and they care about the future of our democracy,” he argued. “That is not academic,” he continued. “That is not theoretical. That is talking about future generations, that’s talking about your happiness, that’s talking about how you see the world. This is big … a paradigm shift.” The leader in question was not an ordained minister, nor even a religious man. His increasingly popular community is not—technically—a church, synagogue, or temple. And the scripture he referenced wasn’t from the Bible. It was Microsoft Encarta vs. Wikipedia— the story of how a movement of selfmotivated volunteers defeated an army of corporate-funded professionals in a crusade to provide information, back in the bygone days of 2009. “If you’re young,” said the preacher, named David Ryan Polgar, “you’ll need to google it.” The rise of the tech ethics congregation. By GREG M. EPSTEIN | Portrait by Matchull Summers SO23-feature_Epstein.religion.indd 32 8/1/23 8:57 AM SO23-feature_Epstein.religion.indd 33 8/1/23 8:57 AM Polgar, 44, is the founder of All Tech Is Human, a nonprofit organization devoted to promoting ethics and responsibility in tech. Founded in 2018, ATIH is based in Manhattan but hosts a growing range of in-person programming—social mixers, mentoring opportunities, career fairs, and job-seeking resources—in several other cities across the US and beyond, reaching thousands. Such numbers would delight most churches. Like other kinds of congregations, ATIH focuses on relationshipbuilding: the staff invests much of its time, for example, in activities like curating its “Responsible Tech Organization” list, which names over 500 companies in which community members can get involved, and growing its responsible-tech talent pool, a list of nearly 1,400 individuals interested in careers in the field. Such programs, ATIH says, bring together many excellent but often disconnected initiatives, all in line with the ATIH mission “to tackle wicked tech & society issues and co-create a tech future aligned with the public interest.” The organization itself doesn’t David Polgar, often get explicitly political with the founder of All Tech Is Human, on stage op-eds or policy advocacy. Rather, at a recent Responsible All Tech Is Human’s underlying Tech Mixer event in strategy is to quickly expand the New York City. “responsible-tech ecosystem.” In other words, its leaders believe there are large numbers of individuals in and around the technology world, often from marginalized backgrounds, who wish tech focused less on profits and more on being a force for ethics and justice. These people will be a powerful force, Polgar believes, if—as the counterculture icon Timothy Leary famously exhorted—they can “find the others.” If that sounds like reluctance to take sides on hot-button issues in tech policy, or to push for change directly, Polgar calls it an “agnostic” business model. And such a model has real strengths, including the ability to bring tech culture’s opposing tribes together under one big tent. But as we’ll see, attempts to stay above the fray can cause more problems than they solve. Meanwhile, All Tech Is Human is growing so fast, with over 5,000 members on its Slack channel as of this writing, that if it were a church, it would soon deserve the prefix “mega.” The group has also consistently impressed me with its inclusiveness: the volunteer and professional leadership of women and people of color is a point of major emphasis, and speaker lineups are among the most heterogeneous I’ve seen in any tech-related endeavor. Crowds, too, are full of young professionals from diverse backgrounds who participate in programs out of passion and curiosity, not hope of financial gain. Well, at least attendees don’t go to ATIH for direct financial gain; as is true with many successful religious congregations, the organization serves as an intentional incubator for professional networking. Still, having interviewed several dozen attendees, I’m convinced that many are hungry for communal support as they navigate a world in which tech has become a transcendent force, for better or worse. Growth has brought things to a turning point. ATIH now stands to receive millions of dollars—including funds from large foundations and tech philanthropist demigods who once ignored it. And Polgar now finds himself in a networking stratosphere with people like Canadian prime minister Justin Trudeau, among other prominent politicos. Will the once-humble community remain dedicated to centering people on the margins of tech culture? SO23-feature_Epstein.religion.indd 34 Or will monied interests make it harder to fight for the people Christian theologians might call “the least of these”? I first started looking into ATIH in late 2021, while researching my forthcoming book Tech Agnostic: How Technology Became the World’s Most Powerful Religion, and Why It Desperately Needs a Reformation (MIT Press, 2024). The book project began because I’d been coming across a striking number of similarities between modern technological culture and religion, and the parallels felt important, given my background. I am a longtime (nonreligious) chaplain at both Harvard and MIT. After two decades immersed in the world of faith, back in 2018 I gave up on what had been my dream: to build a nonprofit “godless congregation” for the growing population of atheists, agnostics, and the religiously unaffiliated. Having started that work just before social media mavens like Mark Zuckerberg began to speak of “connecting the world,” I ultimately lost faith in the notion of building community around either religion or secularism when I realized that technology had overtaken both. Indeed, tech seems to be the dominant force in our economy, politics, and culture, not to mention a daily obsession that can increasingly look like an addiction from which some might plausibly seek the help of a higher power to recover. Tech culture has long been known for its prophets (Jobs, Gates, Musk, et al.), and tech as a whole is even increasingly oriented around moral and ethical messages, such as Google’s infamous “Don’t be evil.” The tech-as-religion comparison I’ve found myself drawing is often unflattering to tech leaders and institutions. Technosolutionism and related ideas can function as a kind of theology, justifying harm in the here and now with the promise of a sweet technological hereafter; powerful CEOs and investors can form the center of a kind of priestly hierarchy, if not an outright caste system; high-tech weapons and surveillance systems seem to threaten an apocalypse of biblical proportions. When I discovered ATIH, I was pleasantly surprised to find a potentially positive example of the sort of dynamic I was describing. COURTESY OF ALL TECH IS HUMAN 34 8/1/23 8:57 AM 35 I am the sort of atheist who admits that certain features of religion can offer people real benefits. And ATIH seemed to be succeeding precisely because it genuinely operated like a secular, techethics-focused version of a religious congregation. “It does work that way,” Polgar acknowledged in February 2022, in the first of our several conversations on the topic. Since then, I’ve continued to admire ATIH’s communal and ethical spirit, while wondering whether communities devoted explicitly to tech ethics might just help bring about a reformation that saves tech from itself. Along with admiration, I’ve also sought to determine whether ATIH is worthy of our faith. recently emerged as dominant and ubiquitous forces across society and culture. Adopting the title “tech ethicist,” he began to write a series of missives on digital health and the idea of “co-creating a better tech future.” His 2017 Medium post “All Tech Is Human,” about how technology design should be informed by more than robotic rationality or utility, generated enthusiastic response and led to the formal founding of the organization a year later. The ATIH concept took a while to catch on, Polgar told me. He worked unpaid for three years and came “close to quitting.” But his background inspired perseverance. Born in 1979 in Cooperstown, New York, Polgar was a philosophical kid who admired Nikola Tesla and wanted to be an inventor. “Why can’t I start something big,” he remembers thinking back then, “even from a little place like this?” Why a congregation? Despite their growing influence, Polgar and the organization discovered ATIH’s events in late 2021, first through the online continue to emphasize their outsider status. ATIH, he argues, Responsible Tech University Summit, a day-long program is building its following in significant part with people who, for dedicated to exploring the intersections of tech ethics and cam- their interest in ethical approaches to technology, feel as unjustly pus life. (One of ATIH’s signature programs is its Responsible ignored as he and many of his upstate peers felt in the shadow Tech University Network, which involves, among other things, a of New York City. ATIH’s model, says the organization’s head of partnerships, growing group of over 80 student “university ambassadors” who represent the organization on their campuses.) All the organiza- Sandra Khalil, is to offer not a “sage on the stage” but, rather, a tion’s programs are organized around typical tech ethics themes, “guide on the side.” Khalil, a veteran of the US Departments of like “the business case for AI ethics,” but State and Homeland Security, also came participants attend as much for the comto the organization with an outsider’s pugTechno-solutionism and munity as for the topic at hand. nacity, feeling “severely underutilized” in Sarah Husain, who’d worked on Twitter’s previous roles as a non-lawyer intent on related ideas can function Trust and Safety team until it was eliminated “challenging the status quo.” as a kind of theology, by Elon Musk, told me at a May 2022 event Polgar, however, hardly shrinks from that several colleagues in her field had spoopportunities to influence tech discourse, justifying harm in the ken highly of ATIH, recommending she whether through media interviews with here and now with attend. Chana Deitsch, an undergraduate outlets like the BBC World News or by joinbusiness student who participates in ATIH’s ing advisory boards like TikTok’s content the promise of a sweet advisory council. ATIH admits, in its “Ten mentorship program, says it not only helps technological hereafter. Principles,” that it draws both from grasswith job leads and reference letters but provides a sense of confidence and belonging. roots models, which it says “have ideas but Alex Sarkissian, formerly a Deloitte conoften lack power,” and from “top-down” sultant and now a Buddhist chaplaincy student, feels that the ones, which can “lack a diversity of ideas” but “have power.” organization has potential “to be a kind of spiritual community The organization does not ask for or accept membership fees for me in addition to my sangha [Buddhist congregation].” from participants, relying instead on major donations solicited I’ve encountered mainly earnest and insightful members like by Polgar and his team, who control decision-making. There these, people who come together for serious mutual support and hasn’t seemed to be a significant call for more democracy—yet. ethical reflection and—non-trivially—fun around a cause I’ve come to hold dear. Granted, few ATIH participants, in my obserThe founder as a god? vation, hold C-level tech positions, which could undermine the organization’s claims that it has the ability to unite stakeholders art of why I’m insisting ATIH is a congregation is that the toward effectual action … or perhaps it simply signifies a popgroup assembled around Polgar demonstrates a religious zeal ulism that could eventually put sympathizers in high places? for organizing and relationship-building as tools for advancing Despite my skepticism toward both theology and technol- positive moral values. Case in point: Rebekah Tweed, ATIH’s ogy, ATIH has often given me the feeling that I’ve found my associate director, once worked in an actual church, as a youth own tech tribe. pastor; now she applies a skill set my field calls “pastoral care” to creating mutually supportive space for ethically minded techies. In 2020, Tweed volunteered on ATIH’s first major public projGrowing pains ect, the Responsible Tech Guide, a crowdsourced document that olgar is a nerdily charismatic former lawyer who has been highlighted the hundreds of people and institutions working in developing the ideas and networks from which the organization the field. After she formally joined the organization, it landed its sprouted for over a decade. As a young professor of business law at first big-time donation: $300,000 over two years from the Ford a couple of small, under-resourced colleges in Connecticut in the Foundation, to pay her salary as well as Polgar’s. They were its early 2010s, he began pondering the ethics of technologies that had first full-time employees. I P P SO23-feature_Epstein.religion.indd 35 8/1/23 8:57 AM 36 SO23-feature_Epstein.religion.indd 36 Digital Sunday school I n September 2022, I attended Building a Better Tech Future for Children, an ATIH event cohosted with the Joan Ganz Cooney Center at Sesame Workshop, a nonprofit research and innovation lab associated with the legendary children’s TV show Sesame Street. This struck me as a shrewd partnership for ATIH: every congregation needs a Sunday school. A community organization aspiring to the advancement of humanity and the betterment of the world will inevitably turn its thoughts to educating the next generation according to its values. After a keynote from Elizabeth Milovidov, senior manager for digital child safety at the Lego Group, on designing digital experiences with children’s well-being in mind came a panel featuring speakers from influential players such as the Omidyar Network and TikTok, as well as young activists. The group discussed the risks and harms facing young people online, and the general tone was optimistic that various efforts to protect them would be successful, particularly if built upon one another. “Digital spaces can be a positive source in the lives of young people,” said the moderator, Mina Aslan. Also on the panel was Harvard Medical School professor Michael Rich, a self-proclaimed “mediatrician”—a portmanteau of “media’’ and “pediatrician.” Rich made good points—for example, stressing the importance of asking kids what they’re hoping for from tech, not just talking about the risks they confront. But one comment triggered my spider-sense: when he said that today’s tech is like his generation’s cigarettes, in that you can’t just tell kids “Don’t do it.” The analogy between tobacco and social media is at best a bizarre one to draw. Millions of young people became smokers not just through peer pressure, but because for decades, Big Tobacco’s whole business model was built on undue corporate influence and even outright lying, including paying influential doctors and scientists to downplay the death they COURTESY OF ALL TECH IS HUMAN Polgar was repeatedly rebuffed in early attempts to recruit large gifts, but of late, the growing ATIH team has received significant support from sources including Melinda French Gates’s Pivotal Ventures and about half a million dollars each from Schmidt Futures (the philanthropic fund of former Google CEO Eric Schmidt) and the Patrick J. McGovern Foundation (yet another tech billionaire’s fortune). The question is: Can an organization that serves a truly inclusive audience, emphasizing humanity and ethics in its own name, afford to get in bed with Fortune 500 companies like Google and Microsoft and/or multibillionaires who will inevitably be motivated by a desire to seem ethical and responsible, even when they decidedly are not? Or rather, can it afford not to do so, when growth means the organization’s staff can grow (and earn a living wage)? And could such tensions someday cause a full-blown schism in the ATIH community? The potential challenges first came to light for me at a May 2022 Not a “sage on the summit in New York. For the first stage” but a “guide on the side”: ATIH head time in several large ATIH events I of partnerships Sandra had personally observed, the meetKhalil moderates an ing featured an invited speaker event in London. employed by one of the world’s largest tech companies: Harsha Bhatlapenumarthy, a governance manager at Meta and also a volunteer leader in a professional association called Trust and Safety. Bhatlapenumarthy—whose panel was called “Tech Policy & Social Media: Where are we headed?”—avoided addressing any of her employer’s recent controversies. Instead of offering any meaningful comment in response to Meta’s troubles over its handling of things from pro-anorexia content to election misinformation, she spoke only vaguely about its ethical responsibilities. The company, she said, was focused on “setting the content moderator up for success.” Which is an interesting way to describe a situation in which Meta had, for example, recently been sued for union busting and human trafficking by content moderators in Kenya. Several attendees were taken aback that Bhatlapenumarthy’s advocacy for her powerful employer went essentially unchallenged during the panel. Among them was Yael Eisenstat, Facebook’s former global head of election integrity operations for political advertising and the summit’s closing speaker. In a fireside chat immediately following the panel in which Bhatlapenumarthy participated, Eisenstat, who’d been a whistleblower against her former employer, eloquently dismissed Bhatlapenumarthy’s non-remarks. “I believe [Meta] doesn’t want this on their platform,” she said, referring to violent and deceptive content, “but they will not touch their business model.” Eisenstat added that she would feel “more encouraged” if companies would stop “holding up the founder as a god.” Eisenstat added to me later, by private message, that “sending a more junior-level employee to speak one-directionally about Meta’s vision of responsible tech is somewhat disingenuous.” In inviting such a speaker, couldn’t ATIH reasonably be understood to be implicated in the offense? If Bhatlapenumarthy’s presence as a seeming mouthpiece for Big Tech talking points had been an isolated incident, I might have ignored it. But a few months later, I found myself wondering if a concerning pattern was emerging. 8/1/23 8:57 AM 37 dealt. Surely ATIH’s leadership would want to avoid any hint that such practices would be acceptable in tech? Tobacco eventually became among the most heavily regulated industries in history, with results including, famously, the US surgeon general’s warnings on tobacco ads and packages. Now the current surgeon general, Vivek Murthy, has warned there is “growing evidence” that social media is “associated with harm to young people’s mental health.” But on the panel (and in his commentary elsewhere), Rich only briefly acknowledged such potential harms, forgoing talk of regulating social media for the idea of cultivating “resilience” in the industry’s millions of young customers. To be clear, I agree with Rich that it is a losing strategy to expect young people to completely abstain from social media. But I fear that tech and our broader society alike are not taking nearly enough ethical responsibility for protecting children from what can be powerful engines of harm. And I was disappointed to see Rich’s relatively sanguine views not only expressed but centered at an ATIH meeting. “I see your concern,” Polgar later told me when I asked him about my apprehensions. Raising his brow with a look of surprise when I wondered aloud whether Rich’s funding sources might have affected the commentary he offered for ATIH’s audience, Polgar made clear he did not agree with all the doctor’s views. He also admitted it is his “worst fear” that his organization might be co-opted by funding opportunities that make it harder “to be a speaker of truth.” “Don’t become a parody of yourself,” he said, seeming to turn the focus of his homily inward. Team human S everal months after the Sesame Workshop event, I attended a crowded mixer at ATIH’s now-regular monthly venue, the Midtown Manhattan offices of the VC firm Betaworks, with a very different kind of speaker: the tech critic Douglas Rushkoff, a freethinker who has often spoken of the need for a kind of secular faith in our common humanity in the face of tech capitalism’s quasi-religious extremism. Polgar is a longtime admirer of his work. How much responsibility? “All tech bros are human,” Rushkoff cracked, launching into an enthusiastically received talk. Fresh off ow much responsibility should a a publicity tour for a book about tech bil“responsible tech” organization like Can an organization that lionaires buying luxury bunkers to escape ATIH take—or not—for inviting speakers serves a truly inclusive with corporate ties, especially when it is not a potential doomsday of their own making, fully open with its audience about such ties? Rushkoff provided a starkly antiauthoritaraudience afford to get How obligated is ATIH to publicly interian contrast to the speakers I’d taken issue in bed with Fortune rogate the conclusions of such speakers? with at the earlier events. Rich’s response to questions I’d asked Ultimately, I don’t know whether ATIH 500 companies and/or after his panel was, essentially, that parents will succeed in its attempts to serve what multibillionaires who will Rushkoff would call “team human” rather ought to channel their energies into making “better choices” around tech, which—conthan becoming an accessory to the overinevitably be motivated by veniently for some of the doctor’s corporate whelming wealth tech can generate by a desire to seem ethical? sponsors—lays the responsibility for chilseeming to make humanity commodifiable dren’s safety on the parents instead of the and, ultimately, redundant. I do, however, tech industry. His lab, I later learned, raised nearly $6 million continue to believe that building a more humane tech future will in 2022, at least partly through grants from Meta, TikTok, and require communal support, because none of us can do it alone. I chose the theme of tech agnosticism for my book in part Amazon. When TikTok CEO Shou Chew testified before the US Congress in March 2023, he cited Rich’s lab—and only Rich’s because I am often reminded that I truly don’t know—and lab—as an example of how TikTok used science and medicine to neither do you—when or where tech’s enormous powers protect minors. Does this represent a conflict of interest—and might actually do the good they purport to do. But I suspect therefore a serious ethical failing on the part of both Rich and we’re going to need a lot more of what Neil Postman’s 1992 ATIH for platforming him? I don’t know. I do worry, though, that book Technopoly, an early exploration of the theme of techthere’s something inhumane in Rich’s emphasis on building kids’ as-religion and a precursor to the techlash, called “loving “resilience” rather than interrogating why they should have to resistance fighters.” While I lack prophetic abilities to know be so resilient against tech in the first place. whether Polgar and co. will help spark such a resistance, the What kind of institution does ATIH want to be? One that pushes potential is genuinely there. In a participatory congregation, back against the powerful, or one that upholds a corporate-friendly one can always worry about co-option, as even Polgar himself version of diversity, allowing its wealthy sponsors to remain com- admits he does; but isn’t it also the responsibility of each of fortable at (almost) all times? As the Gospel of Matthew says, no us to actively help keep our communities accountable to their man (or organization of “humans”) can serve two masters. own ethical values? Asking around ATIH’s network about my concerns, I found Let’s maintain our skepticism, while hoping the ethical tech ambivalence. “I do believe it is possible to do research sponsored congregation gives us continued reason to keep the faith. by companies ethically,” said Justin Hendrix, an occasional ATIH participant and editor of Tech Policy Press, a wonky journal in Greg M. Epstein serves as the humanist chaplain at Harvard University and MIT and as the convener for which academics and others tend to critique established tech ethical life at MIT’s Office of Religious, Spiritual, narratives. “But it is right to scrutinize it for signs of impropriety.” and Ethical Life. H SO23-feature_Epstein.religion.indd 37 8/1/23 8:57 AM 38 SO23-feature_Nelson.lending.indd 38 8/2/23 10:43 AM 39 lenders f o s d e r d Hun ting are protesat the changes nce funder. microfina WHAT HAPPENED TO KIVA Is their s about K trike really much coiva, or about h should e ntrol Americaow internat xpect over thens ional aid ir ? By Mara Kardas-N Illus e tration by SO23-feature_Nelson.lending.indd 39 ls on Andrea D ’Aquino 8/2/23 10:43 AM 40 O ne morning in August 2021, as she had nearly every morning for about a decade, Janice Smith opened her computer and went to Kiva.org, the website of the San Francisco–based nonprofit that helps everyday people make microloans to borrowers around the world. Smith, who lives in Elk River, Minnesota, scrolled through profiles of bakers in Mexico, tailors in Uganda, farmers in Albania. She loved the idea that, one $25 loan at a time, she could fund entrepreneurial ventures and help poor people help themselves. But on this particular morning, Smith noticed something different about Kiva’s website. It was suddenly harder to find key information, such as the estimated interest rate a borrower might be charged—information that had been easily accessible just the day before and felt essential in deciding who to lend to. She showed the page to her husband, Bill, who had also become a devoted Kiva lender. Puzzled, they reached out to other longtime lenders they knew. Together, the Kiva users combed through blog posts, press releases, and tax filings, but they couldn’t find a clear explanation of why the site looked so different. Instead, they learned about even bigger shifts— shifts that shocked them. Kiva connects people in wealthier communities with people in poorer ones through small, crowdfunded loans made to individuals through partner companies and organizations around the world. The individual Kiva lenders earn no interest; money is given to microfinance partners for free, and only the original amount is returned. Once lenders get their money back, they can choose to lend again and again. It’s a model that Kiva hopes will foster a perennial cycle of microfinance lending while requiring only a small outlay from each person. This had been the nonprofit’s bread and butter since its founding in 2005. But now, the Smiths wondered if things were starting to change. The Smiths and their fellow lenders learned that in 2019 the organization had begun charging fees to its lending partners. Kiva had long said it offered zero-interest SO23-feature_Nelson.lending.indd 40 funding to microfinance partners, but the Smiths learned that the recently instituted fees could reach 8%. They also learned about Kiva Capital, a new entity that allows large-scale investors—Google is one—to make big investments in microfinance companies and receive a financial return. The Smiths found this strange: thousands of everyday lenders like them had been offering loans return free for more than a decade. Why should Google now profit off a microfinance investment? The Kiva users noticed that the changes happened as compensation to Kiva’s top employees increased dramatically. In 2020, the CEO took home over $800,000. Combined, Kiva’s top 10 executives made nearly $3.5 million in 2020. In 2021, nearly half of Kiva’s revenue went to staff salaries. Considering all the changes, and the eye-popping executive compensation, “the word that kept coming up was ‘shady,’” Bill Smith told me. “Maybe what they did was legal,” he said, “but it doesn’t seem fully transparent.” He and Janice felt that the organization, which relied mostly on grants and donations to stay afloat, now seemed more focused on how to make money than how to create change. Kiva, on the other hand, says the changes are essential to reaching more borrowers. In an interview about these concerns, Kathy Guis, Kiva’s vice president of investments, told me, “All the decisions that Kiva has made and is now making are in support of our mission to expand financial access.” In 2021, the Smiths and nearly 200 other lenders launched a “lenders’ strike.” More than a dozen concerned lenders (as well as half a dozen Kiva staff members) spoke to me for this article. They have refused to lend another cent through Kiva, or donate to the organization’s operations, until the changes are clarified—and ideally reversed. W hen Kiva was founded in 2005, by Matt Flannery and Jessica Jackley, a worldwide craze for microfinance—sometimes called microcredit—was at its height. The UN had dubbed 2005 the “International Year of Microcredit”; a year later, in 2006, Muhammad Yunus and the Grameen Bank he had founded in the 1980s won the Nobel Peace Prize for creating, in the words of the Nobel Committee, “economic and social development from below.” On a trip to East Africa, Flannery and Jackley had a lightbulb moment: Why not expand microfinance by helping relatively wealthy individuals in places like the US and Europe lend to ves i t u c e x e 0 top 1 ’s a v 0. i 2 K , 0 d 2 e n i n i n b o i l m Co .5 mil 3 $ y l r a enue e v n e r ’s a v i made K half of y l r ries. a a l e a n s , 1 ff 2 a t 0 s 2 In went to 8/2/23 10:43 AM 41 Some lenders were disappointed to learn that loans don’t go directly to the borrowers featured on Kiva’s website. Instead, they are pooled together with others’ contributions and sent to partner institutions to distribute. SO23-feature_Nelson.lending.indd 41 relatively poor businesspeople in places like Tanzania and Kenya? They didn’t think the loans Kiva facilitated should come from grants or donations: the money, they reasoned, would then be limited, and eventually run out. Instead, small loans—as little as $25—would be fully repayable to lenders. Connecting wealthier individuals to poorer ones was the “peer-to-peer” part of Kiva’s model. The second part—the idea that funding would be sourced through the internet via the Kiva.org website—took inspiration from Silicon Valley. Flannery and another Kiva cofounder, Premal Shah, both worked in tech—Flannery for TiVo, Shah for PayPal. Kiva was one of the first crowdfunding platforms, launched ahead of popular sites like GoFundMe. But Kiva is less direct than other crowdfunding sites. Although lenders “choose” borrowers through the website, flipping through profiles of dairy farmers and fruit sellers, money doesn’t go straight to them. Instead, the loans that pass through Kiva are bundled together and sent to one of the partnering microfinance institutions. After someone in the US selects, say, a female borrower in Mongolia, Kiva funds a microfinance organization there, which then lends to a woman who wants to set up a business. Even though the money takes a circuitous route, the premise of lending to an individual proved immensely effective. Stories about Armenian bakers and Moroccan bricklayers helped lenders like the Smiths feel connected to something larger, something with purpose and meaning. And because they got their money back, while the feel-good rewards were high, the stakes were low. “It’s not charity,” the website still emphasizes today. “It’s a loan.” The organization covered its operating expenses with funding from the US government and private foundations and companies, as well as donations from individual lenders, who could add a tip on top of their loan to support Kiva’s costs. This sense of individual connection and the focus on facilitating loans rather than donations was what initially drew Janice Smith. She first heard of microfinance through Bill Clinton’s book Giving, and then again through Oprah Winfrey—Kiva.org was included as one of “Oprah’s Favorite Things” in 2010. Smith was particularly enticed by the idea that she could re-lend the same $25 again and again: “I loved looking through borrower profiles and feeling like I was able to help specific people. Even when I realized that the money was going to a [microfinance lender]”—not directly to a borrower—“it still gave me a feeling of a one-on-one relationship with this person.” Kiva’s easy-to-use website and focus on repayments helped further popularize the idea of small loans to the poor. For many Americans, if they’ve heard of microfinance at all, it’s because they or a friend or family member have lent through the platform. As of 2023, according to a Kiva spokesperson, 2.4 million people from more than 190 countries have done so, ultimately reaching more than 5 million borrowers in 95 countries. The spokesperson also pointed to a 2022 study of 18,000 microfinance customers, 88% of whom said their quality of life had improved since accessing a loan or another financial service. A quarter said the loans and other services had increased their ability to invest and grow their business. B ut Kiva has also long faced criticism, especially when it comes to transparency. There was the obvious issue that the organization suggests a direct connection between Kiva.org users and individual borrowers featured on the site, a connection that does not actually exist. But there were also complaints that the interest rates borrowers pay were not disclosed. Although Kiva initially did not charge fees to the microfinance institutions it funneled money through, the loans to the individual borrowers do include interest. The institutions Kiva partners with use that to cover operational costs and, sometimes, make a profit. Critics were concerned about this lack of disclosure given that interest rates on microfinance loans can reach far into the double digits—for more than a decade, some have even soared above 100%. 8/3/23 8:46 AM 42 (Microlenders and their funders have long argued that interest rates are needed to make funding sustainable.) A Kiva spokesperson stressed that the website now mentions “average cost to borrower,” which is not the interest rate a borrower will pay but a rough approximation. Over the years, Kiva has focused on partnering with “impact-first” microfinance lenders—those that charge low interest rates or focus on loans for specific purposes, such as solar lights or farming. Critics also point to studies showing that microfinance has a limited impact on poverty, despite claims that the loans can be transformative for poor people. For those who remain concerned about microfinance overall, the clean, easy narrative Kiva promotes is a problem. By suggesting that someone like Janice Smith can “make a loan, change a life,” skeptics charge, the organization is effectively whitewashing a troubled industry accused of high-priced loans and harsh collection tactics that have reportedly led to suicides, land grabs, and a connection to child labor and indebted servitude. O ver her years of lending through Kiva.org, Smith followed some of this criticism, but she says she was “sucked in” from her first loan. She was so won over by the mission and the method that she soon became, in her words, a “Kivaholic.” Lenders can choose to join “teams” to lend together, and in 2015 she launched one, called Together for Women. Eventually, the team would include nearly 2,500 Kiva lenders—including one who, she says, put his “whole retirement” into Kiva, totaling “millions of dollars.” Smith soon developed a steady routine. She would open her computer first thing in the morning, scroll through borrowers, and post the profiles of those she considered particularly needy to her growing team, encouraging support from other lenders. In 2020, several years into her “Kivaholicism,” Kiva invited team captains like her to join regular calls with its staff, a way to disseminate information to some of the most active members. At first, these calls were cordial. SO23-feature_Nelson.lending.indd 42 But in 2021, as lenders like Smith noticed changes that concerned them, the tone of some conversations changed. Lenders wanted to know why the information on Kiva’s website seemed less accessible. And then, when they didn’t get a clear answer, they pushed on everything else, too: the fees to microfinance partners, the CEO salaries. In 2021 Smith’s husband, Bill, became captain of a new team calling itself Lenders on Strike, which soon had nearly 200 concerned members. The name sent a clear message: “We’re gonna stop lending until you guys get your act together and address the stuff.” Even though they represented a small fraction of those who had lent through Kiva, the striking members had been involved for years, collectively lending millions of dollars—enough, they thought, to get Kiva’s attention. On the captains’ calls and in letters, the strikers were clear about a top concern: the fees now charged to microfinance institutions Kiva works with. Wouldn’t the fees make the loans more expensive to the borrowers? Individual Kiva.org lenders still expected only their original money back, with no return on top. If the money wasn’t going to them, where exactly would it be going? On one call, the Smiths recall, staffers explained that the fees were a way for Kiva to expand. Revenue from the fees—potentially millions of dollars—would go into Kiva’s overall operating budget, covering everything from new programs to site visits to staff salaries. But on a different call, Kiva’s Kathy Guis acknowledged that the fees could be bad for poor borrowers. The higher cost might be passed down to them; borrowers might see their own interest rates, sometimes already steep, rise even more. When I spoke to Guis in June 2023, she told me those at Kiva “haven’t observed” a rise in borrowers’ rates as a direct result of the fees. Because the organization essentially acts as a middleman, it would be hard to trace this. “Kiva is one among a number of funding sources,” Guis explained—often, in fact, a very small slice of a microlender’s overall funding. “And cost of funds is one among a number of factors that influence borrower pricing.” A Kiva spokesperson said the average fee is 2.53%, with fees of 8% charged on only a handful of “longerterm, high-risk loans.” The strikers weren’t satisfied: it felt deeply unfair to have microfinance lenders, and maybe ultimately borrowers, pay for Kiva’s operations. More broadly, they took issue with new programs the revenue was being spent on. Kiva Capital, the new return-seeking investment arm that Google has participated in, was particularly concerning. Several strikers told me that it seemed strange, if not unethical, for an investor like Google to be able to make money off microfinance loans when everyday Kiva lenders had expected no return for more than a decade—a premise that Kiva had touted as key to its model. A Kiva spokesperson told me investors “are receiving a range of returns well below a commercial investor’s expectations for emerging-market debt investments,” but did not give details. Guis said that thanks in part to Kiva Capital, Kiva “reached 33% more borrowers and deployed 33% more capital in 2021.” Still, the Smiths and other striking lenders saw the program less as an expansion and more as a departure from the Kiva they had been supporting for years. Another key concern, strikers told me, is Kiva US, a separate program that offers zero-interest loans to small businesses domestically. Janice Smith had no fundamental problem with the affordable rates, but she found it odd that an American would be offered 0% interest while borrowers in poorer parts of the world were being charged up to 70%, according to the estimates posted on Kiva’s website. “I don’t see why poor people in Guatemala should basically be subsidizing relatively rich people here in Minnesota,” she told me. Guis disagreed, telling me, “I take issue with the idea that systematically marginalized communities in the US are less deserving.” She said that in 2022, nearly 80% of the businesses that received US loans were “owned by Black, Indigenous, and people of color.” 8/2/23 10:43 AM 43 Someone million dol taking home nearly the ship, nolars a year was steera $25 loans. t the lenders and th ing eir After months of discussions, the strikers and Kiva staff found themselves at loggerheads. “They feel committed to fees as a revenue source, and we feel committed to the fact that it’s inappropriate,” Bill Smith told me. Guis stressed that Kiva had gone through many changes throughout its 18 years—the fees, Kiva Capital, and Kiva US being just a few. “You have to evolve,” she said. T he fees and the returns-oriented Kiva Capital felt strange enough. But what really irked the Lenders on Strike was how much Kiva executives were being paid for overseeing those changes. Lenders wanted to know why, according to Kiva’s tax return, roughly $3.5 million had been spent on executive compensation in 2020—nearly double the amount a few years previously. Bill Smith and others I spoke to saw a strong correlation: at the same time Kiva was finding new ways to make money, Kiva’s leadership was bringing home more cash. The concerned lenders weren’t the only ones to see a connection. Several employees I spoke to pointed to questionable decisions made under the four-year tenure of Neville Crawley, who was named CEO in 2017 and left in 2021. Crawley made approximately $800,000 in 2020, his last full year at the organization, and took home just under $750,000 in 2021, even though he left the position in the middle of the year. When I asked Kathy Guis why Crawley made so much for about six months of work, she said she couldn’t answer but would pass that question along to the board. SO23-feature_Nelson.lending.indd 43 Afterward, I received a written response that did not specifically address CEO compensation, instead noting in part, “As part of Kiva’s commitment to compensation best practices, we conduct regular org-wide compensation fairness research, administer salary surveys, and consult market data from reputable providers.” Chris Tsakalakis, who took over from Crawley, earned more than $350,000 in 2021, for about half a year of work. (His full salary and that of Vishtal Ghotge, his successor and Kiva’s newest CEO, are not yet publicly available in Kiva’s tax filings, nor would Kiva release these numbers to us when we requested them.) In 2021, nearly $20 million of Kiva’s $42 million in revenue went to salaries, benefits, and other compensation. According to the striking lenders, Kiva’s board explained that as a San Francisco– based organization, it needed to attract top talent in a field, and a city, dominated by tech, finance, and nonprofits. The last three CEOs have had a background in business and/or tech; Kiva’s board is stacked with those working at the intersection of tech, business, and finance and headed by Julie Hanna, an early investor in Lyft and other Silicon Valley companies. This was especially necessary, the board argued, as Kiva began to launch new programs like Kiva Capital, as well as Protocol, a blockchainenabled credit bureau launched in Sierra Leone in 2018 and then closed in 2022. The Smiths and other striking lenders didn’t buy the rationale. The leaders of other microlenders—including Kiva partners— make far less. For example, the president and CEO of BRAC USA, a Kiva partner and one of the largest nonprofits in the world, made just over $300,00 in 2020—not only less than what Kiva’s CEO earns, but also below what Kiva’s general counsel, chief investment officer, chief strategy officer, executive vice president of engineering, and chief officer for strategic partnerships were paid in 2021, according to public filings. Julie Hanna, the executive chair of Kiva’s board, made $140,000 for working 10 hours a week in 2021. Premal Shah, one of the founders, took home roughly $320,000 as “senior consultant” in 2020. Even among other nonprofits headquartered in expensive American cities, Kiva’s CEO salary is high. For example, the head of the Sierra Club, based in Oakland, made $500,000 in 2021. Meanwhile, the executive director of Doctors Without Borders USA, based in New York City, had a salary of $237,000 in 2020, the same year that the Kiva top executive made roughly $800,000—despite 2020 revenue of $558 million, compared with Kiva’s $38 million. The striking lenders kept pushing—on calls, in letters, on message boards—and the board kept pushing back. They had given their rationale, about the salaries and all the other changes, and as one Kiva lender told me, it was clear “there would be no more conversation.” Several strikers I spoke to said it was the last straw. This was, they realized, no longer their Kiva. Someone taking home nearly a million dollars a year was steering the ship, not them and their $25 loans. T he Kiva lenders’ strike is concentrated in Europe and North America. But I wanted to understand how the changes, particularly the new fees charged to microfinance lenders, were viewed by the microfinance organizations Kiva works with. So I spoke to Nurhayrah Sadava, CEO of VisionFund Mongolia, who told me she preferred the fees to the old Kiva model. Before the lending fees were introduced, money was lent from Kiva to microfinance organizations in US dollars. The partner organizations then paid the loan back in dollars too. Given high levels of inflation, 8/2/23 10:43 AM 44 site b e w g r .o a the Kiv under, not g n i n g i s e By d tern f s e W e ed h t t a r e o r f c y l i a r v i a K prim wer, o r r o e. b k i r y t a s ’ w s a r r e a f d the the len r o f s n o i t i the cond instability, and currency fluctuations in poorer countries, that meant partners might effectively pay back more than they had taken out. But with the fees, Sadava told me, Kiva now took on the currency risk, with partners paying a little more up front. Sadava saw this as a great deal, even if it looked “shady” to the striking lenders. What’s more, the fees—around 7% to 8% in the case of VisionFund Mongolia—were cheaper than the organization’s other options: their only alternatives were borrowing from microfinance investment funds primarily based in Europe, which charged roughly 20%, or another VisionFund Mongolia lender, which charges the organization 14.5%. Sadava told me that big international donors aren’t interested in funding their microfinance work. Given the context, VisionFund Mongolia was happy with the new arrangement. Sadava says the relatively low cost of capital allowed them to launch “resourcefulness loans” for poor businesswomen, who she says pay 3.4% a month. VisionFund Mongolia’s experience isn’t necessarily representative—it became a Kiva partner after the fees were instituted, and it works in a country where it is particularly difficult to find funding. Still, I was surprised by how resoundingly positive Sadava was about the new model, given the complaints I’d heard from dozens of aggrieved Kiva staffers and lenders. That got me thinking about something Hugh Sinclair, a longtime microfinance staffer and critic, told me a few years back: “The SO23-feature_Nelson.lending.indd 44 client of Kiva is the American who gets to feel good, not the poor person.” I n a way, by designing the Kiva.org website primarily for the Western funder, not the faraway borrower, Kiva created the conditions for the lenders’ strike. For years, Kiva has encouraged the feeling of a personal connection between lenders and borrowers, a sense that through the organization an American can alter the trajectory of a life thousands of miles away. It’s not surprising, then, that the changes at Kiva felt like an affront. (One striker cried when he described how much faith he had put into Kiva, only for Kiva to make changes he saw as morally compromising.) They see Kiva as their baby. So they revolted. Kiva now seems somewhat in limbo. It’s still advertising its old-school, anyonecan-be-a-lender model on Kiva.org, while also making significant operational changes (a private investing arm, the promise of blockchain-enabled technology) that are explicitly inaccessible to everyday Americans—and employing high-flying CEOs with CVs and pedigrees that might feel distant, if not outright off-putting, to them. If Kiva’s core premise has been its accessibility to people like the Smiths, it is now actively undermining that premise, taking a chance that expansion through more complicated means will be better for microfinance than honing the simplistic image it’s been built on. Several of the striking lenders I spoke to were primarily concerned that the Kiva model had been altered into something they no longer recognized. But Janice Smith, and several others, had broader concerns: not just about Kiva, but about the direction the whole microfinance sector was taking. In confronting her own frustrations with Kiva, Smith reflected on criticisms she had previously dismissed. “I think it’s an industry where, depending on who’s running the microfinance institution and the interaction with the borrowers, it can turn into what people call a ‘payday loan’ sort of situation,” she told me. “You don’t want people paying 75% interest and having debt collectors coming after them for the rest of their lives.” Previously, she trusted that she could filter out the most predatory situations through the Kiva website, relying on information like the estimated interest rate to guide her decisions. As information has become harder to come by, she’s had a harder time feeling confident in the terms the borrowers face. In January 2022, Smith closed the 2,500-strong Together for Women group and stopped lending through Kiva. Dozens of other borrowers, her husband included, have done the same. While these defectors represent a tiny fraction of the 2 million people who have used the website, they were some of its most dedicated lenders: of the dozen I spoke to, nearly all had been involved for nearly a decade, some ultimately lending tens of thousands of dollars. For them, the dream of “make a loan, change a life” now feels heartbreakingly unattainable. Smith calls the day she closed her team “one of the saddest days of my life.” Still, the decision felt essential: “I don’t want to be one of those people that’s more like an impact investor who is trying to make money off the backs of the poorer.” “I understand that I’m in the minority here,” she continued. “This is the way [microfinance is] moving. So clearly people feel it’s something that’s acceptable to them, or a good way to invest their money. I just don’t feel like it’s acceptable to me.” Mara Kardas-Nelson is the author of a forthcoming book on the history of microfinance, We Are Not Able to Live in the Sky (Holt, 2024). 8/2/23 10:43 AM Be a part of the biggest conversation happening now. Cut through the AI hype with our weekly newsletter, The Algorithm. Get access to exclusive AI insights and in-depth stories delivered right to your inbox. Scan this code to sign up for free or learn more at technologyreview.com/thealgorithm Untitled-1 1 6/7/23 12:01 PM 46 SO23-feature_Michel.warfare.indd 46 8/2/23 8:37 PM . . . . . . . . . . . . IF A MACHINE TELLS YOU WHEN TO PULL THE TRIGGER, WHO IS ULTIMATELY RESPONSIBLE? . . AIASSISTED WARFARE . . BY . ARTHUR HOLLAND MICHEL . . . . . . . . SO23-feature_Michel.warfare.indd 47 ILLUSTRATIONS . YOSHI SODEOKA . . . . 47 IN a near-future war—one that might begin tomorrow, for all we know—a soldier takes up a shooting position on an empty rooftop. His unit has been fighting through the city block by block. It feels as if enemies could be lying in silent wait behind every corner, ready to rain fire upon their marks the moment they have a shot. Through his gunsight, the soldier scans the windows of a nearby building. He notices fresh laundry hanging from the balconies. Word comes in over the radio that his team is about to move across an open patch of ground below. As they head out, a red bounding box appears in the top left corner of the gunsight. The device’s computer vision system has flagged a potential target—a silhouetted figure in a window is drawing up, it seems, to take a shot. The soldier doesn’t have a clear view, but in his experience the system has a superhuman capacity to pick up the faintest tell of an enemy. So he sets his crosshair upon the box and prepares to squeeze the trigger. In different war, also possibly just over the horizon, a commander stands before a bank of monitors. An alert appears from a chatbot. It brings news that satellites have picked up a truck entering a certain city block that has been designated as a possible staging area for enemy rocket launches. The chatbot has already advised an artillery unit, which it calculates as having the highest estimated “kill probability,” to take aim at the truck and stand by. According to the chatbot, none of the nearby buildings is a civilian structure, though it notes that the determination has yet to be corroborated manually. A drone, which had been dispatched by the system for a closer look, arrives on scene. Its video shows the truck backing into a narrow passage between two compounds. The opportunity to take the shot is rapidly coming to a close. For the commander, everything now falls silent. The chaos, the uncertainty, the cacophony—all reduced to the sound of a ticking clock and the sight of a single glowing button: “APPROVE FIRE ORDER.” 8/2/23 8:37 PM 48 To pull the trigger—or, as the case may be, not to pull it. To hit the button, or to hold off. Legally—and ethically—the role of the soldier’s decision in matters of life and death is preeminent and indispensable. Fundamentally, it is these decisions that define the human act of war. It should be of little surprise, then, that states and civil society have taken up the question of intelligent autonomous weapons— weapons that can select and fire upon targets without any human input—as a matter of serious concern. In May, after close to a decade of discussions, parties to the UN’s Convention on Certain Conventional Weapons agreed, among other recommendations, that militaries using them probably need to “limit the duration, geographical scope, and scale of the operation” to comply with the laws of war. The line was nonbinding, but it was at least an acknowledgment that a human has to play a part—somewhere, sometime—in the immediate process leading up to a killing. But intelligent autonomous weapons that fully displace human decision-making have (likely) yet to see real-world use. Even the “autonomous” drones and ships fielded by the US and other powers are used under close human supervision. Meanwhile, intelligent systems that merely guide the hand that pulls the trigger have been gaining purchase in the warmaker’s tool kit. And they’ve quietly become sophisticated enough to raise novel questions—ones that are trickier to answer than the wellcovered wrangles over killer robots and, with each passing day, more urgent: What does it mean when a decision is only part human and part machine? And when, if ever, is it ethical for that decision to be a decision to kill? SO23-feature_Michel.warfare.indd 48 F or a long time, the idea of supporting a human decision by computerized means wasn’t such a controversial prospect. Retired Air Force lieutenant general Jack Shanahan says the radar on the F4 Phantom fighter jet he flew in the 1980s was a decision aid of sorts. It alerted him to the presence of other aircraft, he told me, so that he could figure out what to do about them. But to say that the crew and the radar were coequal accomplices would be a stretch. That has all begun to change. “What we’re seeing now, at least in the way that I see this, is a transition to a world [in] which you need to have humans and machines … operating in some sort of team,” says Shanahan. The rise of machine learning, in particular, has set off a paradigm shift in how militaries use computers to help shape the crucial decisions of warfare—up to, and including, the ultimate decision. Shanahan was the first director of 8/3/23 10:03 AM 49 Project Maven, a Pentagon program that developed target recognition algorithms for video footage from drones. The project, which kicked off a new era of American military AI, was launched in 2017 after a study concluded that “deep learning algorithms can perform at near-human levels.” (It also sparked controversy—in 2018, more than 3,000 Google employees signed a letter of protest against the company’s involvement in the project.) With machine-learning-based decision tools, “you have more apparent competency, more breadth” than earlier tools afforded, says Matt Turek, deputy director of the Information Innovation Office at the Defense Advanced Research Projects Agency. “And perhaps a tendency, as a result, to turn over more decision-making to them.” A soldier on the lookout for enemy snipers might, for example, do so through the Assault Rifle Combat Application System, a gunsight sold by the Israeli defense firm Elbit Systems. According to a company spec sheet, the “AI-powered” device is capable of “human target detection” at a range of more than 600 yards, and human target “identification” (presumably, discerning whether a person is someone who could be shot) at about the length of a football field. Anna AhronheimCohen, a spokesperson for the company, told MIT Technology Review, “The system has already been tested in real-time scenarios by fighting infantry soldiers.” Another gunsight, built by the company Smartshooter, is advertised as having similar capabilities. According to the company’s website, it can also be packaged into a remote-controlled machine gun like the one that Israeli agents used to assassinate the Iranian nuclear scientist Mohsen Fakhrizadeh in 2021. Decision support tools that sit at a greater remove from the battlefield can be just as decisive. The Pentagon appears to have used AI in the sequence of intelligence analyses and decisions leading up to a potential strike, a process known as a kill chain—though it has been cagey on the details. In response to questions from MIT Technology Review, Laura McAndrews, an Air Force spokesperson, wrote that the service “is utilizing a human-machine teaming approach.” Other countries are more openly experimenting with such automation. Shortly after the IsraelPalestine conflict in 2021, the Israel Defense Forces said it had used The range of judgment calls that go into military decision-making is vast. And it doesn’t always take artificial super-intelligence to dispense with them by automated means. SO23-feature_Michel.warfare.indd 49 what it described as AI tools to alert troops of imminent attacks and to propose targets for operations. The Ukrainian army uses a program, GIS Arta, that pairs each known Russian target on the battlefield with the artillery unit that is, according to the algorithm, best placed to shoot at it. A report by The Times, a British newspaper, likened it to Uber’s algorithm for pairing drivers and riders, noting that it significantly reduces the time between the detection of a target and the moment that target finds itself under a barrage of firepower. Before the Ukrainians had GIS Arta, that process took 20 minutes. Now it reportedly takes one. Russia claims to have its own command-and-control system with what it calls artificial intelligence, but it has shared few technical details. Gregory Allen, the director of the Wadhwani Center for AI and Advanced Technologies and one of the architects of the Pentagon’s current AI policies, told me it’s important to take some of these claims with a pinch of salt. He says some of Russia’s supposed military AI is “stuff that everyone has been doing for decades,” and he calls GIS Arta “just traditional software.” The range of judgment calls that go into military decision-making, however, is vast. And it doesn’t always take artificial superintelligence to dispense with them by automated means. There are tools for predicting enemy troop movements, tools for figuring out how to take out a given target, and tools to estimate how much collateral harm is likely to befall any nearby civilians. None of these contrivances could be called a killer robot. But the technology is not without its perils. Like any complex computer, an AI-based tool might glitch in unusual and unpredictable ways; it’s not clear 8/2/23 8:37 PM 50 that the human involved will always be able to know when the answers on the screen are right or wrong. In their relentless efficiency, these tools may also not leave enough time and space for humans to determine if what they’re doing is legal. In some areas, they could perform at such superhuman levels that something ineffable about the act of war could be lost entirely. Eventually militaries plan to use machine intelligence to stitch many of these individual instruments into a single automated network that links every weapon, commander, and soldier to every other. Not a kill chain, but—as the Pentagon has begun to call it—a kill web. In these webs, it’s not clear whether the human’s decision is, in fact, very much of a decision at all. Rafael, an Israeli defense giant, has already sold one such product, Fire Weaver, to the IDF (it has also demonstrated it to the DoD and the German military). According to company materials, Fire Weaver finds enemy positions, notifies the unit that it calculates as being best placed to fire on them, and even sets a crosshair on the target directly in that unit’s weapon sights. The human’s role, according to one video of the software, is to choose between two buttons: “Approve” and “Abort.” L et’s say that the silhouette in the window was not a soldier, but a child. Imagine that the truck was not delivering warheads to the enemy, but water pails to a home. Of the Department of Defense’s five “ethical principles for artificial intelligence,” which are phrased as qualities, the one that’s always listed first is “Responsible.” In practice, this means that when things go wrong, someone—a human, not a machine—has got to hold the bag. Of course, the principle of responsibility long predates the SO23-feature_Michel.warfare.indd 50 onset of artificially intelligent machines. All the laws and mores of war would be meaningless without the fundamental common understanding that every deliberate act in the fight is always on someone. But with the prospect of computers taking on all manner of sophisticated new roles, the age-old precept has newfound resonance. “Now for me, and for most people I ever knew in uniform, this was core to who we were as commanders, that somebody ultimately will be held responsible,” says Shanahan, who after Maven became the inaugural director of the Pentagon’s Joint Artificial Intelligence Center and oversaw the development of the AI ethical principles. This is why a human hand must squeeze the trigger, why a human hand must click “Approve.” If a computer sets its sights upon the wrong target, and the soldier squeezes the trigger anyway, that’s on the soldier. “If a human does something that leads to an accident with the machine—say, dropping a weapon where it shouldn’t have—that’s still a human’s decision that was made,” Shanahan says. But accidents happen. And this is where things get tricky. Modern militaries have spent hundreds of years figuring out how to differentiate the unavoidable, blameless tragedies of warfare from acts of malign intent, misdirected fury, or gross negligence. Even now, this remains a difficult task. Outsourcing a part of human agency and judgment to algorithms built, in many cases, around the mathematical principle of optimization will challenge all this law and doctrine in a fundamentally new way, says Courtney Bowman, global director of privacy and civil liberties engineering at Palantir, a US-headquartered firm that builds data management software for Of the Department of Defense’s 5 “ethical principles for artificial intelligence,” which are phrased as QUALITIES, the one that’s always listed first is “RESPONSIBLE.” militaries, governments, and large companies. “It’s a rupture. It’s disruptive,” Bowman says. “It requires a new ethical construct to be able to make sound decisions.” This year, in a move that was inevitable in the age of ChatGPT, Palantir announced that it is developing software called the Artificial Intelligence Platform, which allows for the integration of large language models into the company’s military products. In a demo of AIP posted to YouTube this spring, the platform alerts the user to a potentially threatening enemy movement. It then suggests that a drone be sent for a closer look, proposes three possible plans to intercept the offending force, and maps out an optimal route for the selected attack team to reach them. And yet even with a machine capable of such apparent cleverness, militaries won’t want the user to blindly trust its every suggestion. If the human presses only one button in a kill chain, it probably should not be the “I believe” button, as a concerned but anonymous Army operative once put it in a DoD war game in 2019. In a program called Urban Reconnaissance through Supervised Autonomy (URSA), DARPA built a system that enabled robots and drones to act as forward observers for platoons in urban operations. After input from the project’s advisory group on ethical and legal issues, it was decided that the software would only ever designate people as “persons of interest.” Even though the purpose of the technology was to help root out ambushes, it would never go so far as to label anyone as a “threat.” This, it was hoped, would stop a soldier from jumping to the wrong conclusion. It also had a legal rationale, according to Brian Williams, an adjunct research staff member at the 8/2/23 8:37 PM 51 “If people of interest are identified on a screen as red dots, that’s going to have a different subconscious implication than if people of interest are identified on a screen as little happy faces.” Institute for Defense Analyses who led the advisory group. No court had positively asserted that a machine could legally designate a person a threat, he says. (Then again, he adds, no court had specifically found that it would be illegal, either, and he acknowledges that not all military operators would necessarily share his group’s cautious reading of the law.) According to Williams, DARPA initially wanted URSA to be able to autonomously discern a person’s intent; this feature too was scrapped at the group’s urging. Bowman says Palantir’s approach is to work “engineered inefficiencies” into “points in the decisionmaking process where you actually do want to slow things down.” For example, a computer’s output that points to an enemy troop movement, he says, might require a user to seek out a second corroborating source of intelligence before proceeding with an action (in the video, the Artificial Intelligence Platform does not appear to do this). In the case of AIP, Bowman says that the idea is to present the information in such a way “that the viewer understands, the analyst understands, this is only a suggestion.” In practice, protecting human judgment from the sway SO23-feature_Michel.warfare.indd 51 of a beguilingly smart machine could come down to small details of graphic design. “If people of interest are identified on a screen as red dots, that’s going to have a different subconscious implication than if people of interest are identified on a screen as little happy faces,” says Rebecca Crootof, a law professor at the University of Richmond, who has written extensively about the challenges of accountability in human-in-the-loop autonomous weapons. In some settings, however, soldiers might only want an “I believe” button. Originally, DARPA envisioned URSA as a wrist-worn device for soldiers on the front lines. “In the very first working group meeting, we said that’s not advisable,” Williams told me. The kind of engineered inefficiency necessary for responsible use just wouldn’t be practicable for users who have bullets whizzing by their ears. Instead, they built a computer system that sits with a dedicated operator, far behind the action. But some decision support systems are definitely designed for the kind of split-second decisionmaking that happens right in the thick of it. The US Army has said that it has managed, in live tests, to shorten its own 20-minute targeting cycle to 20 seconds. Nor does the market seem to have embraced this same spirit of restraint. In demo videos posted online, the bounding boxes for the computerized gunsights of both Elbit and Smartshooter are blood red. O ther times, the computer will be right and the human will be wrong. If the soldier on the rooftop had second-guessed the gunsight, and it turned out that the silhouette was in fact an enemy sniper, his teammates could have paid a heavy price for his split second of hesitation. This is a different source of trouble, much less discussed but no less likely in real-world combat. And it puts the human in something of a pickle. Soldiers will be told to treat their digital assistants with enough mistrust to safeguard the sanctity of their judgment. But with machines that are often right, this same reluctance to defer to the computer can itself become a point of avertable failure. Aviation history has no shortage of cases where a human pilot’s refusal to heed the machine led to catastrophe. These (usually perished) souls have not been looked upon kindly by investigators seeking to explain the tragedy. Carol J. Smith, a senior research scientist at Carnegie Mellon University’s Software Engineering Institute who helped craft responsible AI guidelines for the DoD’s Defense Innovation Unit, doesn’t see an issue: “If the person in that moment feels that the decision is wrong, they’re making it their call, and they’re going to have to face the consequences.” For others, this is a wicked ethical conundrum. The scholar M.C. Elish has suggested that a human who is placed in this kind of impossible loop could end up serving as what she calls a “moral crumple 8/2/23 8:37 PM 52 zone.” In the event of an accident—regardless of whether the human was wrong, the computer was wrong, or they were wrong together—the person who made the “decision” will absorb the blame and protect everyone else along the chain of command from the full impact of accountability. In an essay, Smith wrote that the “lowest-paid person” should not be “saddled with this responsibility,” and neither should “the highest-paid person.” Instead, she told me, the responsibility should be spread among everyone involved, and that the introduction of AI should not change anything about that responsibility. In practice, this is harder than it sounds. Crootof points out that even today, “there’s not a whole lot of responsibility for accidents in war.” As AI tools become larger and more complex, and as kill chains become shorter and more web-like, finding the right people to blame is going to become an even more labyrinthine task. Those who write these tools, and the companies they work for, aren’t likely to take the fall. Building AI software is a lengthy, iterative process, often drawing from opensource code, which stands at a distant remove from the actual material facts of metal piercing flesh. And barring any significant changes to US law, defense contractors are generally protected from liability anyway, says Crootof. Any bid for accountability at the upper rungs of command, meanwhile, would likely find itself stymied by the heavy veil of government classification that tends to cloak most AI decision support tools and the manner in which they are used. The US Air Force has not been forthcoming about whether its AI has even seen real-world use. Shanahan says Maven’s AI models were deployed SO23-feature_Michel.warfare.indd 52 for intelligence analysis soon after the project launched, and in 2021 the secretary of the Air Force said that “AI algorithms” had recently been applied “for the first time to a live operational kill chain,” with an Air Force spokesperson at the time adding that these tools were available in intelligence centers across the globe “whenever needed.” But Laura McAndrews, the Air Force spokesperson, said that in fact these algorithms “were not applied in a live, operational kill chain” and declined to detail any other algorithms that may, or may not, have been used since. The real story might remain shrouded for years. In 2018, the Pentagon issued a determination that exempts Project Maven from Freedom of Information requests. Last year, it handed the entire program to the National Geospatial-Intelligence Agency, which is responsible for processing America’s vast intake of secret aerial surveillance. Responding to questions about whether the algorithms are used in kill chains, Robbin Brooks, an NGA spokesperson, told MIT Technology Review, “We can’t speak to specifics of how and where Maven is used.” I n one sense, what’s new here is also old. We routinely place our safety—indeed, our entire existence as a species—in the hands of other people. Those decision-makers defer, in turn, to machines that they do not entirely comprehend. In an exquisite essay on automation published in 2018, at a time when operational AI-enabled decision support was still a rarity, former Navy secretary Richard Danzig pointed out that if a president “decides” to order a nuclear strike, it will not be because anyone has looked out the window of the Oval Office and seen enemy missiles raining down on DC but, rather, because those missiles have been detected, tracked, and identified—one hopes correctly—by algorithms in the air defense network. As in the case of a commander who calls in an artillery strike on the advice of a chatbot, or a rifleman who pulls the trigger at the mere sight of a red bounding box, “the most that can be said is that ‘a human being is involved,’” Danzig wrote. “This is a common situation in the modern age,” he wrote. “Human decisionmakers are riders traveling across obscured terrain with little or no ability to assess the powerful beasts that carry and guide them.” As AI tools become larger and more complex, and as kill chains become shorter and more web-like, finding the right people to blame is going to become an even more labyrinthine task. 8/2/23 8:37 PM 53 There can be an alarming streak of defeatism among the people responsible for making sure that these beasts don’t end up eating us. During a number of conversations I had while reporting this story, my interlocutor would land on a sobering note of acquiescence to the perpetual inevitability of death and destruction that, while tragic, cannot be pinned on any single human. War is messy, technologies fail in unpredictable ways, and that’s just that. SO23-feature_Michel.warfare.indd 53 “In warfighting,” says Bowman of Palantir, “[in] the application of any technology, let alone AI, there is some degree of harm that you’re trying to—that you have to accept, and the game is risk reduction.” It is possible, though not yet demonstrated, that bringing artificial intelligence to battle may mean fewer civilian casualties, as advocates often claim. But there could be a hidden cost to irrevocably conjoining human judgment and mathematical reasoning in those ultimate moments of war—a cost that extends beyond a simple, utilitarian bottom line. Maybe something just cannot be right, should not be right, about choosing the time and manner in which a person dies the way you hail a ride from Uber. To a machine, this might be suboptimal logic. But for certain humans, that’s the point. “One of the aspects of judgment, as a human capacity, is that it’s done in an open world,” says Lucy Suchman, a professor emerita of anthropology at Lancaster University, who has been writing about the quandaries of human-machine interaction for four decades. The parameters of life-and-death decisions—knowing the meaning of the fresh laundry hanging from a window while also wanting your teammates not to die—are “irreducibly qualitative,” she says. The chaos and the noise and the uncertainty, the weight of what is right and what is wrong in the midst of all that fury—not a whit of this can be defined in algorithmic terms. In matters of life and death, there is no computationally perfect outcome. “And that’s where the moral responsibility comes from,” she says. “You’re making a judgment.” The gunsight never pulls the trigger. The chatbot never pushes the button. But each time a machine takes on a new role that reduces the irreducible, we may be stepping a little closer to the moment when the act of killing is altogether more machine than human, when ethics becomes a formula and responsibility becomes little more than an abstraction. If we agree that we don’t want to let the machines take us all the way there, sooner or later we will have to ask ourselves: Where is the line? Arthur Holland Michel writes about technology. He is based in Barcelona and can be found, occasionally, in New York. 8/2/23 8:37 PM 54 From supersize slideshows to Steve Jobs’s Apple keynote, corporate presentations have always pushed technology forward. By Claire L. Evans Above: To celebrate the launch of the 1987 Saab 9000 CD sedan, an audience of 2,500 was treated to an hourlong operetta involving 26-foot-tall projection screens, a massive chorus, the entire Stockholm Philharmonic, and some 50 performers. The greatest slideshow on Earth SO23-back_presentations.indd 54 7/31/23 2:39 PM 55 It’s 1948, and it isn’t a great year for alcohol. Prohibition has come and gone, and booze is a buyer’s market again. That much is obvious from Seagram’s annual sales meeting, an 11-city traveling extravaganza designed to drum up nationwide sales. No expense has been spared: there’s the twohour, professionally acted stage play about the life of a whiskey salesman. The beautiful anteroom displays. The free drinks. But the real highlight is a slideshow. To call the Seagram-Vitarama a slideshow is an understatement. It’s an experience: hundreds of images of the distilling process, set to music, projected across five 40-by-15-foot screens. “It is composed of pictures, yet it is not static,” comments one awed witness. “The overall effect is one of magnificence.” Inspired by an Eastman Kodak exhibit at the 1939 World’s Fair, the SO23-back_presentations.indd 55 Seagram-Vitarama is the first A/V presentation ever given at a sales meeting. It will not be the last. In the late ’40s, multimedia was a novelty. But by the early 1960s, nearly all companies with national advertising budgets were using multimedia gear—16-millimeter projectors, slide projectors, filmstrip projectors, and overheads—in their sales training and promotions, for public relations, and as part of their internal communications. Many employed in-house A/V directors, who were as much showmen as technicians. Because although presentations have a reputation for being tedious, when they’re done right, they’re theater. The business world knows it. Ever since the days of the Vitarama, companies have leveraged the dramatic power of images to sell their ideas to the world. Next slide, please The sound of slides clacking is deafening. But it doesn’t matter, because the champagne is flowing and the sound system is loud. The 2,500 dignitaries and VIPs in the audience are being treated to an hourlong operetta about luxury travel. Onstage, a massive chorus, the entire Stockholm Philharmonic, and some 50 dancers and performers are fluttering around a pair of Saab 9000CD sedans. Stunning images of chrome details, leather seats, and open roads dance across a 26-foot-tall screen behind them. The images here are all analog: nearly 7,000 film slides, carefully arranged in a grid of 80 Kodak projectors. It’s 1987, and slideshows will never get any bigger than this. Before PowerPoint, and long before digital projectors, 35-millimeter film slides were king. Bigger, clearer, and less expensive to 7/31/23 2:39 PM 56 produce than 16-millimeter film, and more colorful and higher-resolution than video, slides were the only medium for the kinds of high-impact presentations given by CEOs and top brass at annual meetings for stockholders, employees, and salespeople. Known in the business as “multi-image” shows, these presentations required a small army of producers, photographers, and live production staff to pull off. First the entire show had to be written, storyboarded, and scored. Images were selected from a library, photo shoots arranged, animations and special effects produced. A white-gloved technician developed, mounted, and dusted each slide before dropping it into the carousel. Thousands of cues were programmed into the show control computers—then tested, and tested again. Because computers crash. Projector bulbs burn out. Slide carousels get jammed. SO23-back_presentations.indd 56 “When you think of all the machines, all the connections, all the different bits and pieces, it’s a miracle these things even played at all,” says Douglas Mesney, a commercial photographer turned slide producer whose company Incredible Slidemakers produced the 80-projector Saab launch. Now 77 years old, he’s made a retirement project of archiving the now-forgotten slide business. Mesney pivoted to producing multi-image shows in the early 1970s after an encounter with an impressive six-screen setup at the 1972 New York Boat Show. He’d been shooting spreads for Penthouse and car magazines, occasionally lugging a Kodak projector or two to pitch meetings for advertising clients. “All of a sudden you look at six projectors and what they can do, and you go, Holy mackerel,” he remembers. Six was just the beginning. At the height of Mesney's career, his shows called for up to 100 projectors braced together in vertiginous rigs. With multiple projectors pointing toward the same screen, he could create seamless panoramas and complex animations, all synchronized to tape. Although the risk of disaster was always high, when he pulled it off, his shows dazzled audiences and made corporate suits look like giants. Mesney’s clients included IKEA, Saab, Kodak, and Shell; he commanded production budgets in the hundreds of thousands of dollars. And in the multi-image business, that was cheap. Larger A/V staging companies, like Carabiner International, charged up to $1 million to orchestrate corporate meetings, jazzing up their generic multiimage “modules” with laser light shows, dance numbers, and top-shelf talent like THIS SPREAD & PREVIOUS: DOUGLAS MESNEY/INCREDIBLE SLIDEMAKERS “All of a sudden you look at six projectors and what they can do, and you go, Holy mackerel.” 7/31/23 2:39 PM 57 Douglas Mesney (above), a former commercial photographer, produced shows with production budgets in the hundreds of thousands of dollars for clients including IKEA, Saab, Kodak, and Shell. Hall & Oates, the Allman Brothers, and even the Muppets. “I liken it to being a rock-and-roll roadie, but I never went on the tour bus,” explains Susan Buckland, a slide programmer who spent most of her career behind the screen at Carabiner. From its incorporation in 1976 to the mid-1980s, the Association for Multi-Image, a trade association for slide producers, grew from zero to 5,000 members. At its peak, the multi-image business employed some 20,000 people and supported several festivals and four different trade magazines. One of these ran a glowing profile of Douglas Mesney in 1980; when asked for his prognosis about the future of slides, he replied: “We could make a fortune or be out of business in a year.” He wasn’t wrong. At the time, some 30 manufacturers of electronic slide programming devices SO23-back_presentations.indd 57 vied for the multi-image dollar. To meet the demand for high-impact shows, the tech had quickly evolved from manual dissolve units and basic control systems—programmed with punched paper tape, and then audiocassette—to dedicated slide control computers like the AVL Eagle I, which could drive 30 projectors at once. The Eagle, which came with word processing and accounting software, was a true business computer—so much so that when Eagle spun off from its parent company, Audio Visual Labs, in the early ’80s, it became one of Silicon Valley’s most promising computer startups. Eagle went public in the summer of 1983, making its president, Dennis R. Barnhart, an instant multimillionaire. Only hours after the IPO, Barnhart plowed his brand-new cherry-red Ferrari through a guardrail near the company’s headquarters in Los Gatos, California, flipped through the air, crashed into a ravine, and died. The slide business would soon follow. Douglas Mesney likes to say that if you never saw a slide show, you never will. The machines to show them have been landfilled. The slides themselves were rarely archived. Occasionally a few boxes containing an old multi-image “module” will turn up in a storage unit, and occasionally those will even be undamaged. But with the exception of a few hobbyists and retired programmers, the know-how to restore and stage multi-image slideshows is scarce. This leaves former slide professionals at a loss. “All of us are devastated that none of the modules survived,” says Susan Buckland. “Basically, I don’t have a past, because I can’t explain it.” The entire industry, which existed at an unexpected intersection of 7/31/23 2:39 PM It wasn’t long before the computers that ran the slide shows evolved beyond the medium. analog and high-tech artistry, came and went in a little over 20 years. Presentations, like porn, have always pushed technology forward; in the multiimage days, producers like Mesney took the slide as far as it could go, using every tool available to create bigger and bolder shows. Mesney claims to have set the land speed record for a slide presentation with a three-minute-long, 2,400-slide show, but even at top speed, slides are static. The computers that controlled them, however, were not—and it wasn’t long before they evolved beyond the medium. “Back then, computers were fast enough to tell slides what to do, but they weren’t fast enough to actually create the images themselves,” explains Steven Michelsen, a former slide programmer who restores and runs old multi-image shows in his Delaware garage. “It took another 10 or SO23-back_presentations.indd 58 15 years until you could run a show straight from your computer and have the images look worth looking at,” he adds. The last slide projector ever made rolled off the assembly line in 2004. The inside of its casing was signed by factory workers and Kodak brass before the unit was handed over to the Smithsonian. Toasts and speeches were made, but by then they were eulogies, because PowerPoint had already eaten the world. Inventing PowerPoint The Hotel Regina is an Art Nouveau marvel overlooking the Tuileries Garden and the Louvre. But on this day in 1992, its Old World meeting rooms have been retrofitted with advanced video technology. The color projector in the back of the room, the size of a small refrigerator, cost upwards of $100,000 and takes an hour to warm up. A team of technicians has spent the better part of the last 48 hours troubleshooting to ensure that nothing goes wrong when Robert Gaskins, the fastidious architect of a new piece of software called PowerPoint 3.0, walks into the room. He’ll be carrying a laptop under his arm, and when he reaches the lectern, he’ll pick up a video cable, plug it in, and demonstrate for the first time something that has been reproduced billions of times since: a video presentation, running straight off a laptop, in full color. The audience, full of Microsoft associates from across Europe, will go bananas. They “grasped immediately what the future would bring for their own presentations,” Gaskins later wrote. “There was deafening applause.” It’s hard now to imagine deafening applause for a PowerPoint—almost as hard TOP ROW: RICHARD SHIPPS/DD&B STUDIO; DOUGLAS MESNEY/INCREDIBLE SLIDEMAKERS; WILDEN ENTERPRISES MIDDLE ROW: DOUGLAS MESNEY/INCREDIBLE SLIDEMAKERS; WILDEN ENTERPRISES; RICHARD SHIPPS/DD&B STUDIOS 58 7/31/23 2:39 PM BOTTOM ROW: WILDEN ENTERPRISES; RICHARD SHIPPS/DD&B STUDIOS; DOUGLAS MESNEY/INCREDIBLE SLIDEMAKERS; IMAGES COURTESY STEVEN MICHELSEN 59 as it is to imagine anyone but Bob Gaskins standing at this particular lectern, ushering in the PowerPoint age. Presentations are in his blood. His father ran an A/V company, and family vacations usually included a trip to the Eastman Kodak factory. During his graduate studies at Berkeley, he tinkered with machine translation and coded computer-generated haiku. He ran away to Silicon Valley to find his fortune before he could finalize his triple PhDs in English, linguistics, and computer science, but he brought with him a deep appreciation for the humanities, staffing his team with like-minded polyglots, including a disproportionately large number of women in technical roles. Because Gaskins ensured that his offices—the only Microsoft division, at the time, in Silicon Valley—housed a museum-worthy art collection, PowerPoint’s SO23-back_presentations.indd 59 architects spent their days among works by Frank Stella, Richard Diebenkorn, and Robert Motherwell. Gaskins’s 1984 proposal for PowerPoint, written when he was VP of product development at the Sunnyvale startup Forethought, is a manifesto in bullet points. It outlines the slumbering, largely-hidden-from-view $3.5 billion business presentation industry and its enormous need for clear, effective slides. It lists technology trends—laser printers, color graphics, “WYSIWYG” software—that point to an emerging desktop presentation market. It’s a stunningly prescient document throughout. But Gaskins italicized only one bullet point in the whole thing. User benefits: Allows the content-originator to control the presentation. This is Gaskins’s key insight: a presentation’s message is inevitably diluted when its production is outsourced. In the early ’80s, he meant that literally. The first two versions of PowerPoint were created to help executives produce their own overhead transparencies and 35-millimeter slides, rather than passing the job off to their secretaries or a slide bureau. “In the ’50s, ’60s, and early ’70s, information flow was narrow,” explains Sandy Beetner, former CEO of Genigraphics, a business graphics company that was, for several decades, the industry leader in professional presentation graphics. Their clients were primarily Fortune 500 companies and government agencies with the resources to produce full-color charts, 3D renderings, and other high-tech imagery on those slides. Everyone else was limited to acetate 7/31/23 2:39 PM 60 With multiple projectors pointing toward the same screen, producers could create seamless panoramas and complex animations, all synchronized to tape. SO23-back_presentations.indd 60 PowerPoint had become shorthand for the stupefying indignities of office life— a 2001 New Yorker profile summed it up as “software you impose on other people.” the States 10 years later, an expert in antique concertinas. By then, PowerPoint had become shorthand for the stupefying indignities of office life. A 2001 New Yorker profile summed it up as “software you impose on other people”; the statistician Edward Tufte, known for his elegant monographs about data visualization, famously blamed the 2003 Columbia shuttle disaster on a bum PowerPoint slide. Gaskins’s software, Tufte argued, produces relentlessly sequential, hierarchical, sloganeering, over-managed presentations, rife with “chartjunk” and devoid of real meaning. No wonder software corporations loved it. Robert Gaskins is remarkably sympathetic to these views, not least because Tufte’s mother, the Renaissance scholar Virginia Tufte, mentored him as an undergraduate in the English department at the University of Southern California. In a reflection written on the 20th anniversary of PowerPoint’s introduction, Gaskins acknowledged that “more business and academic talks look like poor attempts at sales presentations,” a phenomenon he blamed as much on a “mass failure of taste” WILDEN ENTERPRISES overheads and—gasp—words. “Prior to PowerPoint,” she says, “people communicated in black and white. There was just so much missed in that environment.” Beetner oversaw Genigraphics’ national network service bureaus, which were located in every major American city and staffed 24 hours a day, 365 days a year, by graphic artists prepared to produce, polish, and print slides. The company was so vital to presentational culture that Gaskins negotiated a deal to make Genigraphics the official 35-millimeter slide production service for PowerPoint 2.0; a “Send to Genigraphics” menu command was baked into PowerPoint until 2003. This, incidentally, was around the same time that Kodak stopped making Carousel projectors. Gaskins retired from Microsoft in 1993 and moved to London. He returned to 7/31/23 2:39 PM 61 61 as on PowerPoint itself, a tool so powerful it collapsed all preexisting contexts. Not everything’s a sales presentation; nor should it be. But PowerPoint made it easy to add multimedia effects to informal talks, empowering lay users to make stylistic decisions once reserved for professionals. To paraphrase an early PowerPoint print ad: now the person making the presentation made the presentation. That those people weren’t always particularly good at it didn’t seem to matter. What did matter was that presentations were no longer reserved for yearend meetings and big ideas worthy of the effort and expense required to prepare color slides. “The scalability of information and audience that PowerPoint brought to the party was pretty incredible,” says Beetner, whose company has survived SO23-back_presentations.indd 61 as a ghost in the machine, in the form of PowerPoint templates and clip art. “It opened up the channels dramatically, and pretty quickly. There isn’t a student alive, at any level, that hasn’t seen a PowerPoint presentation.” Indeed, PowerPoint is used in religious sermons; by schoolchildren preparing book reports; at funerals and weddings. In 2010, Microsoft announced that PowerPoint was installed on more than a billion computers worldwide. At this scale, PowerPoint’s impact on how the world communicates has been immeasurable. But here’s something that can be measured: Microsoft grew tenfold in the years that Robert Gaskins ran its Graphics Business Unit, and it has grown 15-fold since. Technology corporations, like PowerPoint itself, have exploded. And so have their big presentations, which are no longer held behind closed doors. They’re now semi-public affairs, watched—willingly and enthusiastically— by consumers around the world. Nobody has to worry about slide carousels getting jammed anymore, but things still go haywire all the time, from buggy tech demos to poorly-thought-out theatrics. When everything works, a good presentation can drive markets and forge reputations. Of course, this particular evolution wasn’t exclusively Microsoft’s doing. Because perhaps the most memorable corporate presentation of all time— Steve Jobs’s announcement of the iPhone at Macworld 2007— wasn’t a PowerPoint at all. It was a Keynote. Claire L. Evans is a writer and musician exploring ecology, technology, and culture. 7/31/23 2:39 PM 62 SO23-back_opensource.indd 62 8/1/23 3:22 PM 63 40 Open source at Free and open-source software are now foundational to modern code, but much about them is still in flux. By Rebecca Ackermann Illustration by Saiman Chow SO23-back_opensource.indd 63 When Xerox donated a new laser printer to the MIT Artificial Intelligence Lab in 1980, the company couldn’t have known that the machine would ignite a revolution. The printer jammed. And according to the 2002 book Free as in Freedom, Richard M. Stallman, then a 27-year-old programmer at MIT, tried to dig into the code to fix it. He expected to be able to: he’d done it with previous printers. The early decades of software development generally ran on a culture of open access and free exchange, where engineers could dive into each other’s code across time zones and institutions to make it their own or squash a few bugs. But this new printer ran on inaccessible proprietary software. Stallman was locked out—and enraged that Xerox had violated the open code-sharing system he’d come to rely on. A few years later, in September 1983, Stallman released GNU, an operating system designed to be a free alternative to one of the dominant operating systems at the time: Unix. Stallman envisioned GNU as a means to fight back against the proprietary mechanisms, like copyright, that were beginning to flood the tech industry. The free-software movement was born from one frustrated engineer’s simple, rigid philosophy: for the good of the world, all code should be open, without restriction or commercial intervention. Forty years later, tech companies are making billions on proprietary software, and much of the technology around us— from ChatGPT to smart thermostats—is inscrutable to everyday consumers. In this environment, Stallman’s movement may look like a failed values experiment crushed under the weight of commercial reality. But in 2023, the free and open-source software movement is not only alive and well; it has become a keystone of the tech industry. Today, 96% of all code bases incorporate open-source software. GitHub, the biggest platform for the open-source community, is used by more than 100 million developers worldwide. The Biden administration’s Securing Open Source Software Act of 2022 publicly recognized open-source software as critical economic and security infrastructure. Even AWS, Amazon’s money-making cloud arm, supports the development and 8/1/23 3:22 PM 64 maintenance of open-source software; it committed its portfolio of patents to an open use community in December of last year. Over the last two years, while public trust in private technology companies has plummeted, organizations including Google, Spotify, the Ford Foundation, Bloomberg, and NASA have established new funding for open-source projects and their counterparts in open science efforts—an extension of the same values applied to scientific research. The fact that open-source software is now so essential means that long-standing leadership and diversity issues in the movement have become everyone’s problems. Many open-source projects began with “benevolent dictator for life” (BDFL) models of governance, where original founders hang on to leadership for years—and not always responsibly. Stallman and some other BDFLs have been criticized by their own communities for misogynistic or even abusive behavior. Stallman stepped down as president of the Free Software Foundation in 2019 (although he returned to the board two years later). Overall, open-source participants are still overwhelmingly white, male, and located in the Global North. Projects can be overly influenced by corporate interests. Meanwhile, the people doing the hard work of keeping critical code healthy are not consistently funded. In fact, many major open-source projects still operate almost completely on volunteer steam. Challenges notwithstanding, there’s plenty to celebrate in 2023, the year of GNU’s 40th birthday. The modern opensource movement persists as a collaborative haven for transparent ways of working within a highly fragmented and competitive industry. Selena Deckelmann, chief product and technology officer at the Wikimedia Foundation, says the power of open source lies in its “idea that people anywhere can collaborate together on software, but also on many [more] things.” She points out that tools to put this philosophy into action, like mailing lists, online chat, and open version control systems, were pioneered in opensource communities and have been adopted as standard practice by the wider tech industry. “We found a way for people from all SO23-back_opensource.indd 64 over the world, regardless of background, to find a common cause to collaborate with each other,” says Kelsey Hightower, an early contributor to Kubernetes, an open-source system for automating app deployment and management, who recently retired from his role as a distinguished engineer at Google Cloud. “I think that is pretty unique to the world of open source.” The 2010s backlash against tech’s unfettered growth, and the recent AI boom, have focused a spotlight on the open-source movement’s ideas about who has the right to use other people’s information online and who benefits from technology. Clement Delangue, CEO of the open-source AI company Hugging Face, which was recently valued at $4 billion, testified before Congress modified versions too. Stallman saw free software as an essential right: “Free as in free speech, not free beer,” as his apocryphal slogan goes. He created the GNU General Public License, what’s known as a “copyleft” license, to ensure that the four freedoms were protected in code built with GNU. Linus Torvalds, the Finnish engineer who in 1991 created the now ubiquitous Unix alternative Linux, didn’t buy into this dogma. Torvalds and others, including Microsoft’s Bill Gates, believed that the culture of open exchange among engineers could coexist with commerce, and that more-restrictive licenses could forge a path toward both financial sustainability and protections for software creators and users. It was during a 1998 strategic meeting of free-software “If a company only ends up just sharing, and nothing more, I think that should be celebrated.” in June of 2023 that “ethical openness” in AI development could help make organizations more compliant and transparent, while allowing researchers beyond a few large tech companies access to technology and progress. “We’re in a unique cultural moment,” says Danielle Robinson, executive director of Code for Science and Society, a nonprofit that provides funding and support for public-interest technology. “People are more aware than ever of how capitalism has been influencing what technologies get built, and whether you have a choice to interact with it.” Once again, free and open-source software have become a natural home for the debate about how technology should be. Free as in freedom The early days of the free-software movement were fraught with arguments about the meaning of “free.” Stallman and the Free Software Foundation (FSF), founded in 1985, held firm to the idea of four freedoms: people should be allowed to run a program for any purpose, study how it works from the source code and change it to meet their needs, redistribute copies, and distribute advocates—which notably did not include Stallman—that this pragmatic approach became known as “open source.” (The term was coined and introduced to the group not by an engineer, but by the futurist and nanotechnology scholar Christine Peterson.) Karen Sandler, executive director of the Software Freedom Conservancy, a nonprofit that advocates for free and open-source software, saw firsthand how the culture shifted from orthodoxy to a big-tent approach with room for for-profit entities when she worked as general counsel at the Software Freedom Law Center in the early 2000s. “The people who were ideological—some of them stayed quite ideological. But many of them realized, oh, wait a minute, we can get jobs doing this. We can do well by doing good,” Sandler remembers. By leveraging the jobs and support that early tech companies were offering, open-source contributors could sustain their efforts and even make a living doing what they believed in. In that manner, companies using and contributing to free and open software could expand the community beyond volunteer enthusiasts and improve the work itself. “How could we 8/2/23 2:15 PM 65 PETER ADAMS Christine Peterson, a futurist and lecturer in the field of nanotechnology, coined the term “open source” in 1998. ever make it better if it’s just a few radical people?” Sandler says. As the tech industry grew around private companies like Sun Microsystems, IBM, Microsoft, and Apple in the late ’90s and early ’00s, new open-source projects sprang up, and established ones grew roots. Apache emerged as an open-source web server in 1995. Red Hat, a company offering enterprise companies support for open-source software like Linux, went public in 1999. GitHub, a platform originally created to support version control for open-source projects, launched in 2008, the same year that Google released Android, the first opensource phone operating system. The more pragmatic definition of the concept came to dominate the field. Meanwhile, Stallman’s original philosophy persisted among dedicated groups of believers—where it still lives today through nonprofits like FSF, which only uses and advocates for software that protects the four freedoms. As open-source software spread, a bifurcation of the tech stack became standard practice, with open-source code as the support structure for proprietary work. Free and SO23-back_opensource.indd 65 open-source software often served in the underlying foundation or back-end architecture of a product, while companies vigorously pursued and defended copyrights on the user-facing layers. Some estimate that Amazon’s 1999 patent on its one-click buying process was worth $2.4 billion per year to the company until it expired. It relied on Java, an open-source programming language, and other open-source software and tooling to build and maintain it. Today, corporations not only depend on open-source software but play an enormous role in funding and developing open-source projects: Kubernetes (initially launched and maintained at Google) and Meta’s React are both robust sets of software that began as internal solutions freely shared with the larger technology community. But some people, like the Software Freedom Conservancy’s Karen Sandler, identify an ongoing conflict between profitdriven corporations and the public interest. “Companies have become so savvy and educated with respect to open-source software that they use a ton of it. That’s good,” says Sandler. At the same time, they profit from their proprietary work—which they sometimes attempt to pass off as open too, a practice the scholar and organizer Michelle Thorne dubbed “openwashing” in 2009. For Sandler, if companies don’t also make efforts to support user and creator rights, they’re not pushing forward the free and open-source ethos. And she says for the most part, that’s indeed not happening: “They’re not interested in giving the public any appreciable rights to their software.” Others, including Kelsey Hightower, are more sanguine about corporate involvement. “If a company only ends up just sharing, and nothing more, I think that should be celebrated,” he says. “Then if for the next two years you allow your paid employees to work on it, maintaining the bugs and issues, but then down the road it’s no longer a priority and you choose to step back, I think we should thank [the company] for those years of contributions.” In stark contrast, FSF, now in its 38th year, holds firm to its original ideals and opposes any product or company that does not support the ability for users to view, modify, and redistribute code. The group today runs public action campaigns like “End Software Patents,” publishing articles and submitting amicus briefs advocating the end of patents on software. The foundation’s executive director, Zoë Kooyman, hopes to continue pushing the conversation toward freedom rather than commercial concerns. “Every belief system or form of advocacy needs a far end,” she says. “That’s the only way to be able to drive the needle. [At FSF], we are that far end of the spectrum, and we take that role very seriously.” Free as in puppy Forty years on from the release of GNU, there is no singular open-source community, “any more than there is an ‘urban community,’” as researcher and engineer Nadia Asparouhova (formerly Eghbal) writes in her 2020 book Working in Public: The Making and Maintenance of Open Source Software. There’s no singular definition, either. The Open Source Initiative (OSI) was founded in 1998 to steward the meaning of the phrase, but not all modern open-source projects 8/2/23 3:00 PM 66 adhere to the 10 specific criteria OSI laid out, and other definitions appear across communities. Scale, technology, social norms, and funding also range widely from project to project and community to community. For example, Kubernetes has a robust, organized community of tens of thousands of contributors and years of Google investment. Salmon is a niche open-source bioinformatics research tool with fewer than 50 contributors, supported by grants. OpenSSL, which encrypts an estimated 66% of the web, is currently maintained by 18 engineers compensated through donations and elective corporate contracts. The major discussions now are more about people than technology: What does healthy and diverse collaboration look like? How can those who support the code get what they need to continue the work? “How do you include a voice for all the people affected by the technology you build?” asks James Vasile, an open-source consultant and strategist who sits on the board of the Electronic Frontier Foundation. “These are big questions. We’ve never grappled with them before. No one was working on this 20 years ago, because that just wasn’t part of the scene. Now it is, and we [in the open-source community] have the chance to consider these questions.” “Free as in puppy,” a phrase that can be traced back to 2006, has emerged as a valuable definition of “free” for modern open-source projects—one that speaks to the responsibilities of creators and users to each other and the software, in addition to their rights. Puppies need food and care to survive; open-source code needs funding and “maintainers,” individuals who consistently respond to requests and feedback from a community, fix bugs, and manage the growth and scope of a project. Many opensource projects have become too big, complicated, or important to be governed by one person or even a small group of like-minded individuals. And open-source contributors have their own needs and concerns, too. A person who’s good at building may not be good at maintaining; someone who creates a project may not want to or be able to run it indefinitely. In 2018, for instance, Guido SO23-back_opensource.indd 66 van Rossum, the creator of the open-source programming language Python, stepped down from leadership after almost 30 years, exhausted from the demands of the mostly uncompensated role. “I’m tired,” he wrote in his resignation message to the community, “and need a very long break.” Supporting the people who create, maintain, and use free and open-source software requires new roles and perspectives. Whereas the movement in its early days was populated almost exclusively by engineers communicating across message boards and through code, today’s open-source projects invite participation from new disciplines to handle logistical work like growth and advocacy, as well as efforts toward greater inclusion and belonging. “We’ve shifted inclusion strategy, took that responsibility very seriously. To find out where things stood, the company partnered with the Linux Foundation in 2021 on a survey and resulting report on diversity and inclusion within open source. The data showed that despite a pervasive ethos of collaboration and openness (more than 80% of the respondents reported feeling welcome), communities are dominated by contributors who are straight, white, male, and from the Global North. In response, Cheatham, who is now the company’s chief of staff, focused on ways to broaden access and promote a sense of belonging. GitHub launched All In for Students, a mentorship and education program with 30 students drawn primarily from historically Black colleges and universities. “We need designers, ethnographers, social and cultural experts. We need everyone to be playing a role in open source.” from open source being about just the technical stuff to the broader set of expertise and perspectives that are required to make effective open-source projects,” says Michael Brennan, senior program officer with the Technology and Society program at the Ford Foundation, which funds research into open internet issues. “We need designers, ethnographers, social and cultural experts. We need everyone to be playing a role in open source if it’s going to be effective and meet the needs of the people around the world.” One powerful source of support arrived in 2008 with the launch of GitHub. While it began as a version control tool, it has grown into a suite of services, standards, and systems that is now the “highway system” for most open-source development, as Asparouhova puts it in Working in Public. GitHub helped lower the barrier to entry, drawing wider contribution and spreading best practices such as community codes of conduct. But its success has also given a single platform vast influence over communities dedicated to decentralized collaboration. Demetris Cheatham, until recently GitHub’s senior director for diversity and In its second year, the program expanded to more than 400 students. Representation has not been the only stumbling block to a more equitable opensource ecosystem. The Linux Foundation report showed that only 14% of opensource contributors surveyed were getting paid for their work. While this volunteer spirit aligns with the original vision of free software as a commerce-free exchange of ideas, free labor presents a major access issue. Additionally, 30% of respondents in the survey did not trust that codes of conduct would be enforced—suggesting they did not feel they could count on a respectful working environment. “We’re at another inflection point now where codes of conduct are great, but they’re only a tool,” says Code for Science and Society’s Danielle Robinson. “I’m starting to see larger cultural shifts toward rethinking extractive processes that have been a part of open source for a long time.” Getting maintainers paid and connecting contributors with support are now key to opening up open source to a more diverse group of participants. 8/2/23 3:00 PM 67 With that in mind, this year GitHub established resources specifically for maintainers, including workshops and a hub of DEI tools. And in May, the platform launched a new project to connect large, well-resourced open-source communities with smaller ones that need help. Cheatham says it’s crucial to the success of any of these programs that they be shared for free with the broader community. “We’re not inventing anything new at all. We’re just applying open-source principles to diversity, equity, and inclusion,” she says. GitHub’s influence over open source may be large, but it is not the only group working to get maintainers paid and expand opensource participation. The Software Freedom Conservancy’s Outreachy diversity initiative offers paid internships; as of 2019, 92% of past Outreachy interns have identified as women and 64% as people of color. Opensource fundraising platforms like Open Collective and Tidelift have also emerged to help maintainers tap into resources. The philanthropic world is stepping up too. The Ford Foundation, the Sloan Foundation, Omidyar Network, and the Chan Zuckerberg Initiative, as well as smaller organizations like Code for Science and Society, have all recently begun or expanded their efforts to support open-source research, contributors, and projects—including specific efforts promoting inclusion and diversity. Govind Shivkumar from Omidyar Network told MIT Technology Review that philanthropy is well positioned to establish funding architecture that could help prove out open-source projects, making them less risky prospects for future governmental funding. In fact, research supported by the Ford Foundation’s Digital Infrastructure Fund contributed to Germany’s recent creation of a national fund for open digital infrastructure. Momentum has also been building in the US. In 2016 the White House began requiring at least 20% of governmentdeveloped software to be open source. Last year’s Securing Open Source Software Act passed with bipartisan support, establishing a framework for attention and investment at the federal level toward making opensource software stronger and more secure. SO23-back_opensource.indd 67 The fast-approaching future Open source contributes valuable practices and tools, but it may also offer a competitive advantage over proprietary efforts. A document leaked in May from Google argued that open-source communities had pushed, tested, integrated, and expanded the capabilities of large language models more thoroughly than private efforts could’ve accomplished on their own: “Many of the new ideas [in AI development] are from ordinary people. The barrier to entry for training and experimentation has dropped from the total output of a major research organization to one person, an evening, and a beefy laptop.” The recently articulated concept of Time till Open Source Alternative (TTOSA)—the time between the release of a proprietary product and an open-source equivalent—also speaks to this advantage. One researcher estimated the average TTOSA to be seven years but noted that the process has been speeding up thanks to easy-to-use services like GitHub. At the same time, much of our modern world now relies on underfunded and rapidly expanding digital infrastructure. There has long been an assumption within open source that bugs can be identified and solved quickly by the “many eyes” of a wide community—and indeed this can be true. But when open-source software affects millions of users and its maintenance is handled by handfuls of underpaid individuals, the weight can be too much for the system to bear. In 2021, a security vulnerability in a popular open-source Apache library exposed an estimated hundreds of millions of devices to hacking attacks. Major players across the industry were affected, and large parts of the internet went down. The vulnerability’s lasting impact is hard to quantify even now. Other risks emerge from open-source development without the support of ethical guardrails. Proprietary efforts like Google’s Bard and OpenAI’s ChatGPT have demonstrated that AI can perpetuate existing biases and may even cause harm—while also not providing the transparency that could help a larger community audit the technology, improve it, and learn from its mistakes. But allowing anyone to use, modify, and distribute AI models and technology could accelerate their misuse. One week after Meta began granting access to its AI model LLaMA, the package leaked onto 4chan, a platform known for spreading misinformation. LLaMA 2, a new model released in July, is fully open to the public, but the company has not disclosed its training data as is typical in open-source projects—putting it somewhere in between open and closed by some definitions, but decidedly not open by OSI’s. (OpenAI is reportedly working on an open-source model as well but has not made a formal announcement.) “There are always trade-offs in the decisions you make in technology,” says Margaret Mitchell, chief ethics scientist at Hugging Face. “I can’t just be wholeheartedly supportive of open source in all cases without any nuances or caveats.” Mitchell and her team have been working on opensource tools to help communities safeguard their work, such as gating mechanisms to allow collaboration only at the project owner’s discretion, and “model cards” that detail a model’s potential biases and social impacts—information researchers and the public can take into consideration when choosing which models to work with. Open-source software has come a long way since its rebellious roots. But carrying it forward and making it into a movement that fully reflects the values of openness, reciprocity, and access will require careful consideration, financial and community investment, and the movement’s characteristic process of self-improvement through collaboration. As the modern world becomes more dispersed and diverse, the skill sets required to work asynchronously with different groups of people and technologies toward a common goal are only growing more essential. At this rate, 40 years from now technology might look more open than ever—and the world may be better for it. Rebecca Ackermann is a writer, designer, and artist based in San Francisco. 8/1/23 3:22 PM 68 Researchers can now coax cells like those in this photomicrograph of endometrial tissue into microcosms of the human uterus. SO23-back_organoids.indd 68 7/31/23 4:04 PM 69 Tiny faux organs could finally crack the mystery of menstruation Organoids are helping researchers explore one of the last frontiers of human physiology. By Saima Sidik In the center of the laboratory dish, there was a subtle white film that could only be seen when the light hit the right way. Ayse Nihan Kilinc, a reproductive biologist, popped the dish under the microscope, and an image appeared on the attached screen. As she focused the microscope, the film resolved into clusters of droplet-like spheres with translucent interiors and thin black boundaries. In this magnified view, the structures ranged in size from as small as a quarter to as large as a golf ball. In reality, each was only as big as a few grains of sand. “They’re growing,” Kilinc said, observing that their plump shapes were a promising sign. “These are good organoids.” Kilinc, who works in the lab of biological engineer Linda Griffith at MIT, is among a small group of scientists using new tools akin to miniature organs to study a poorly understood—and frequently SO23-back_organoids.indd 69 problematic—part of human physiology: menstruation. Heavy, sometimes debilitating periods strike at least a third of people who menstruate at some point in their lives, causing some to miss weeks of work or school every year and jeopardizing their professional standing. Anemia threatens about two-thirds of people with heavy periods. And when menstrual blood flows through the fallopian tubes and into the body cavity, it’s thought to sometimes create painful lesions—characteristics of a disease called endometriosis, which can require multiple surgeries to control. No one is entirely sure how—or why— the human body choreographs this monthly dance of cellular birth, maturation, and death. Many people desperately need treatments to make their period more manageable, but it’s difficult for scientists to design medications without understanding how menstruation really works. 7/31/23 4:04 PM 70 An uncommon problem Periods are rare in the animal kingdom. The human body goes through the menstrual cycle to prepare the uterus to welcome a fetus, whether one is likely to show up or not. In contrast, most animals prepare the uterus only once a fetus is already present. That cycle is a constant pattern of wounding and repair. The process starts when levels of a hormone called progesterone plummet, indicating that no baby will be growing in the uterus that month. Removing progesterone triggers a response similar to what happens when the body fights off an infection. Inflammation injures the endometrium. Over the next five or so days, the damaged tissue sloughs off and flows out of the body. As soon as the bleeding starts, the endometrium begins to heal. Over the course of about 10 days, this tissue quadruples in thickness. No other human tissue is known to grow so extensively and so quickly—“not even aggressive cancer cells,” says Jan Brosens, an obstetrician and gynecologist at the University of Warwick in the UK. As the tissue heals—in a rare example of scarless repair—it becomes an environment that can shield an embryo, which is a foreign entity in the body, from an immune system trained to reject interlopers. SO23-back_organoids.indd 70 With a long snout reminiscent of an elephant’s trunk and a body similar to an opossum’s, the elephant shrew was already an oddball when van der Horst learned that it’s one of the few animals that get a period. Scientists have filled in the rough outline of this process after decades of research, but many details remain opaque. How exactly the endometrium repairs itself so extensively is unknown. Why some people have much heavier periods than others remains an open question. And why humans menstruate, rather than reabsorbing unused endometrial tissue like many other mammals, is a matter of hot debate among biologists. This lack of understanding hampers scientists, who would like to find treatments for periods that are too painful to be tamed by over-the-counter painkillers or too heavy to be absorbed by pads and tampons. As a result, many people suffer. A study performed in the Netherlands found that on average women lost about a week of productivity per year because of abdominal pain and other symptoms related to their periods. “It would not be unusual for a patient to see me in the clinic and say that every month, they had to have two or three days off work,” says Hilary Critchley, a gynecologist and reproductive biologist at the University of Edinburgh. Heavy periods can make even daily tasks difficult. Getting up from a chair, for example, can be an ordeal for someone worried about the possibility of having stained the seat. Mothers with low iron levels tend to have babies with low birth weights and other health problems, so the effects of heavy menstruation trickle down through generations. And yet the uterus often goes unacknowledged, even by researchers who are exploring topics like tissue regeneration, to which the organ is clearly relevant, Brosens says. “It is almost unforgivable, in my view,” he adds. Ask researchers why menstruation remains so enigmatic and you’ll get a variety of answers. Most everyone agrees there’s not enough funding to attract the number of researchers the field deserves— as is often the case for health problems that primarily affect women. The fact that menstruation is shrouded in taboos doesn’t help. But some researchers say it has been hard to find the right tools to study the phenomenon. Scientists tend to start studies of the human body in other organisms, such as mice, fruit flies, and yeast, before translating the knowledge back to humans. These so-called “model systems” reproduce quickly and can be altered genetically, and scientists can work with them without running into as many ethical or logistical concerns as they would if they experimented on people. But because menstruation is so rare in the animal kingdom, it’s been tough to find ways to study the process outside the human body. “I think that the main limitations are PREVIOUS SPREAD: GETTY IMAGES That understanding could be in the works, thanks to endometrial organoids— biomedical tools made from bits of the tissue that lines the uterus, called the endometrium. To make endometrial organoids, scientists collect cells from a human volunteer and let those cells self-organize in laboratory dishes, where they develop into miniature versions of the tissue they came from. The research is still very much in its infancy. But organoids have already provided insights into how endometrial cells communicate and coordinate, and why menstruation is routine for some people and fraught for others. Some researchers are hopeful that these early results mark the dawn of a new era. “I think it’s going to revolutionize the way we think about reproductive health,” says Juan Gnecco, a reproductive engineer at Tufts University. 7/31/23 4:04 PM 71 model systems, honestly,” says Julie Kim, a reproductive biologist at Northwestern University. Early adventures “ORGANOID CO-CULTURE MODEL OF THE CYCLING HUMAN ENDOMETRIUM IN A FULLY-DEFINED SYNTHETIC 2 EXTRACELLULAR MATRIX REVEALS EPITHELIAL-STROMAL CROSSTALK.” JUAN S. GNECCO ET AL. In the 1940s, the Dutch zoologist Cornelius Jan van der Horst was among the first scientists to work on an animal model for studying menstruation. Van der Horst was fascinated by unusual, poorly studied critters, and this fascination led him to South Africa, where he trapped and studied the elephant shrew. With a long snout reminiscent of an elephant’s trunk and a body similar to an opossum’s, the elephant shrew was already an oddball when van der Horst learned that it’s one of the few animals that get a period—a fact bats, which live primarily in Central and South America, were not easily accessible, so for several decades his discovery remained simply a point of interest in the scientific literature. Then, in the 1960s, an eager graduate student named John J. Rasweiler IV enrolled at Cornell University. Rasweiler wanted to study a type of animal reproduction that mirrors what happens in humans, so his mentor pointed out Hamlett’s discovery. Perhaps Rasweiler would like to go find some bats and see what he could do with them? “It was a very challenging undertaking,” Rasweiler says. “Essentially I had to invent everything from start to finish.” First there were the trips to Trinidad and Researchers can track how organoids respond to various stimuli. Here endometrial tissue thickens when exposed to a synthetic version of the hormone progesterone, mirroring the lead-up to menstruation. he probably discovered “more or less by accident,” says Anthony Carter, a developmental biologist at the University of Southern Denmark who wrote a review of van der Horst’s work. Elephant shrews are not cooperative study subjects, however. They only menstruate at certain times of year, and they don’t do well in captivity. There’s also the challenge of catching them, which van der Horst and his colleagues attempted with hand-held nets. The shrews were agile, so it was “sometimes a fascinating but mostly a disappointing sport,” he wrote. Around the same time, George W.D. Hamlett, a Harvard-based biologist, discovered an alternative. Hamlett was examining preserved samples of a nectar-loving bat called Glossophaga soricina when he noticed evidence of menstruation. The SO23-back_organoids.indd 71 Colombia to collect the bats. Then there was the issue of how to transport them back to the United States without their getting crushed or overheating. (Shipping them in takeout food containers, bundled together into a larger package, turned out to work well.) Once the bats were in the lab, he had to figure out how to work with them without letting them escape. He ended up constructing a walk-in cage on wheels that he could roll up to the bats’ enclosures. “I loved working with them—delightful animals,” says Rasweiler, who has since retired from a career as a reproductive physiologist at SUNY Downstate. But other researchers were put off by the idea of working with a flying animal. In 2016, the spiny mouse—a rodent that thrives in the dry conditions of the Middle East, South Asia, and parts of Africa—joined the exclusive club of animals known to menstruate. Spiny mice can be raised in the lab, so they may become valuable subjects for menstruation research. But millions of years of evolution lie between humans and mice, leading Brosens to think the genetics underlying their uteruses are likely to differ substantially. Much of the foundational work on menstruation has been performed in macaque monkeys. But primates are expensive to care for, and the Animal Welfare Act places restrictions on primate research that do not apply to other common lab animals. Through a series of manipulations, scientists also found that they could force a common lab mouse to have something similar to a period. This model has been useful, but it’s still only an artificial representation of true human menstruation. What researchers really needed was a way to use humans as study subjects for menstruation research. But even setting aside the obvious ethical concerns, such a thing would be very challenging logistically. The endometrium evolves exceedingly quickly—“at an hourly rate, we see different responses from the cells, different functions,” says Aleksandra Tsolova, a cell biologist at the University of Calgary. “It’s very dynamic tissue.” Researchers would need to perform invasive biopsies almost constantly to study it inside the human body, and even then, altering it genetically or through chemical treatments would be largely impossible. But by the early 1900s, a solution to this problem had already started to emerge. And it was not a creature from the jungle or the African grasslands that paved the road, but an organism from the bottom of the sea. Organoids come on the scene The groundwork for what would become modern-day organoids was laid in 1910, when a zoologist named Henry Van Peters Wilson realized that cells from marine sponges have a sort of “memory” of how they’re arranged in the animal, even after they’re separated. When he dissociated a 8/1/23 1:34 PM 72 sponge by squeezing it through a mesh and then mixed the cells together again, the original sponge re-formed. Midcentury work showed that certain cells from chick embryos have a similar ability. In 2009, a study published in the journal Nature described a possible way of extending these observations to human organs. The researchers took a single adult stem cell from a mouse intestine—which had the ability to become any type of intestinal cell—and embedded it in a gelatinous substance. The cell divided and, together with its progeny, formed a miniature, simplified version of the intestinal lining. It was the first time scientists had laid out a method of creating an organoid from human tissue that was accessible to many labs and straightforward to adapt to other organs. Since then, scientists have extended this general approach to mimic aspects of around a dozen human tissue types, including those from the gut, the kidneys, and the brain—and, by the late 2010s, the uterus. It was a happy accident that brought endometrial organoids into the mix. In the years leading up to their development, scientists had been trying to study the endometrium by growing its cells in smooth layers on the bottoms of laboratory dishes. Stromal cells, which provide structural support for the tissue and play a key role in pregnancy, proved easy to grow this way—these cells secrete a substance that sticks them to each other, and also makes them adhere to petri dishes. But epithelial cells, another critical component of the endometrium, posed a problem. In a dish, they stopped responding to hormones, and their shapes were unlike what’s seen in the human body. Then, while working with a mix of human placental and endometrial tissue in an effort to get the placenta to form organoids, a reproductive biologist named Margherita Turco noticed something serendipitous. If they were suspended in a gel instead of being grown in liquid, and given just the right mix of molecules from the human body, endometrial epithelial cells assembled into tiny three-dimensional simulacra of the organ they came from. SO23-back_organoids.indd 72 “It’s mind-blowing that we are very, very close to the patient, but we’re not working within the patient. There’s huge potential.” “They grew really, really well,” Turco says. In fact, endometrial organoids were “kind of overtaking the cultures.” Another group independently published similar findings around the same time. Today, placental and endometrial organoids are both valuable tools in the lab Turco runs at the Friedrich Miescher Institute for Biomedical Research in Basel, Switzerland. Her original 2017 publication calls for using tissue from a biopsy, rather than stem cells, to make organoids from the endometrium. Some labs instead use tissue removed from people who have had hysterectomies. But Turco’s lab recently showed that bits of the endometrium found in menstrual blood also work, which would mean the new endometrial organoids can be grown without requiring biopsies or surgery. From all these starting points, researchers can now create microcosms of the human uterus. Each organoid reminds Tsolova of a tiny bubble suspended in a gelatinous dessert. And each presents a unique opportunity to understand processes that science has long ignored. Period in a dish Endometrial organoids became integral to the work of the small community of researchers focused on the uterus. Since 2017, many labs have put their own spins on these new tools. Kim’s lab has added stromal cells to the epithelial cells that make up classic endometrial organoids. She and her colleagues mix the two together and simply let the combination “do its thing,” she says. The result is like a malt ball with stromal cells on the inside and epithelial cells on the outside. In 2021, Brosens and his colleagues created similar structures, which they call “assembloids.” Instead of mixing the two cell types together, they created an organoid out of epithelial cells and then added a layer of stromal cells on top. Using assembloids, they’ve learned that deteriorating cells play a key role in helping the embryo implant in the uterus. Because the endometrium is constantly dying and regrowing, the tissue is highly flexible and able to adjust its shape, Brosens explains. This helps the tissue kick-start pregnancy: “Maternal cells will grab the embryo,” he says, “and literally pull that embryo into the tissue.” A video from one of Brosens’s recent publications shows an assembloid remodeling around a five-day-old embryo. Before he and his colleagues did this work, conventional wisdom said the endometrium was passive tissue that was invaded by the embryo, but that’s “just completely wrong,” he says. This new understanding of how embryos implant could improve in vitro 7/31/23 4:04 PM 73 “MENSTRUAL FLOW AS A NON-INVASIVE SOURCE OF ENDOMETRIAL ORGANOIDS.” TEREZA CINDROVA-DAVIES ET AL. COMMUNICATIONS BIOLOGY. fertilization and help explain why some people are prone to miscarriages. Eventually, Critchley hopes, scientists can design treatments that let people choose when to have a period—or if they even want to have one at all. Hormonal birth control can accomplish these goals for some, but these drugs can also cause unscheduled bleeding that makes periods harder to manage, and some people find the side effects of the medication intolerable. To create better options, scientists still need to understand how a normal period works. Making an organoid menstruate in a dish would be a huge boon for achieving this goal, so that’s what some researchers are trying to do. which is a hallmark of the lesions that characterize endometriosis. IL-1β caused organoids to grow rapidly, but only when stromal cells were mixed in along with the epithelial cells. This suggests that signals from stromal cells might be part of what causes endometriosis to develop into a painful condition. Meanwhile, Kilinc is trying to understand why some people’s periods are so heavy. Endometrial tissue growing into the muscle that lines the uterus seems to cause lesions, which can be one source of excessive bleeding. To see how such lesions could form, Kilinc watches how endometrial organoids react when they hit a dense gel, which mimics the texture of muscle. 200 um Margherita Turco's laboratory at the Friedrich Miescher Institute for Biomedical Research in Switzerland has found that organoids derived directly from the endometrium (left) and from menstrual blood (right) of the same person have indistinguishable shapes and structures. By manually adding hormones to organoids, Gnecco and his collaborators can replicate some of what the endometrium experiences over the course of a month. As the cycle progresses, they see the cells adjusting the complement of genes they use, just as they would in the human body. The shape of the organoid also follows a familiar pattern. Glands—infoldings of cells from which mucus and other substances are secreted—change from smooth tubes to sawtooth-like structures as this faux menstrual cycle progresses. With this system working, the next step is to figure out what happens when the endometrium malfunctions. “That’s what really got me excited,” Gnecco says. As a first step, he treated organoids with an inflammatory molecule called IL-1β, SO23-back_organoids.indd 73 In a soft gel, endometrial organoids maintain a nice, round structure. But when the organoid is in a stiff gel, it’s a different story. A video from one of Kilinc’s recent experiments shows an organoid pulsating and squirming, almost like a pot of water that’s about to boil over. Finally, a group of cells shoots off, creating an appendagelike structure that punctures the stiff gel. Videos like this make Kilinc think that contact with muscle might be among the triggers that cause the endometrium to start wounding this tissue and causing heavy bleeding. “But,” she adds, “this is not clear yet—we are still investigating.” Speedier science Today’s endometrial organoids can’t do everything animal models can do. For one thing, they don’t yet include key components of menstruation, like blood vessels and immune cells. For another, they can’t reveal how distant parts of the body, like the brain, influence what happens in the uterus. But because they’re derived from human tissue, they’re intimately related to the bizarre, idiosyncratic process that is a human period, and that’s worth a lot. “It’s mind-blowing that we are very, very close to the patient, but we’re not working within the patient,” Tsolova says. “There’s huge potential.” In parallel to the work on organoids, scientists have created an “organ on a chip” that mimics the endometrium. Tiny tubes affixed to a flat surface carry liquids to endometrial tissue, mimicking the flow of blood or hormones transmitted from other parts of the body. An ideal model system could combine endometrial cells in their natural arrangement—as in an organoid—with flowing liquids, as on a chip. Already, organoids have helped researchers solve old puzzles. Researchers in Vienna, for example, used this technology to figure out which genes cause some endometrial cells to grow cilia—hair-like structures that beat in coordination to move liquid, mucus, and embryos within the uterus. Other researchers have used organoids to learn how endometrial cells mature throughout the menstrual cycle. Meanwhile, Kim and her colleagues used organoids to study how the endometrium responds to abnormal hormone levels, which may be a factor in endometrial cancer. People who menstruate have waited a long time for researchers to tackle such questions. Burdensome periods are often seen as just a “women’s problem”—a mindset Tsolova disagrees with because it ignores the fact that people struggling with menstruation often can’t contribute their full range of talents to their communities. “It’s a societal problem,” she says. “It affects every person, in every way.” Saima Sidik is a freelance science journalist based in Somerville, Massachusetts. 7/31/23 4:04 PM 74 35 Innovators Under 35 Tips for aspiring innovators on trying, failing, and the future of AI. By Andrew Ng How to be an innovator I nnovation is a powerful engine for uplifting society and fueling economic growth. Antibiotics, electric lights, refrigerators, airplanes, smartphones—we have these things because innovators created something that didn’t exist before. MIT Technology Review’s Innovators Under 35 list celebrates individuals who have accomplished a lot early in their careers and are likely to accomplish much more still. Having spent many years working on AI research and building AI products, I’m fortunate to have participated in a few innovations that made an impact, like using reinforcement learning to fly helicopter drones at Stanford, starting and leading Google Brain to drive large-scale deep learning, and creating online courses that led to the founding of Coursera. I’d like to share some thoughts about how to do it well, sidestep some of the pitfalls, and avoid building things that lead to serious harm along the way. SO23-back_35.indd 74 AI is a dominant driver of innovation today As I have said before, I believe AI is the new electricity. Electricity revolutionized all industries and changed our way of life, and AI is doing the same. It’s reaching into every industry and discipline, and it’s yielding advances that help multitudes of people. AI—like electricity—is a generalpurpose technology. Many innovations, such as a medical treatment, space rocket, or battery design, are fit for one purpose. In contrast, AI is useful for generating art, serving web pages that are relevant to a search query, optimizing shipping routes to save fuel, helping cars avoid collisions, and much more. The advance of AI creates opportunities for everyone in all corners of the economy to explore whether or how it applies to their area. Thus, learning about AI creates disproportionately many opportunities to do something that no one else has ever done before. For more than 20 years, this publication has highlighted the work of young innovators through our 35 Innovators Under 35 competition—partly to call attention to what’s going on now, but even more to reveal where technology is headed in the near future. This year we’re excited to include an introductory essay by Andrew Ng (a 35 Innovators honoree himself in 2008) and a profile of our Innovator of the Year, Sharon Li (page 76). To see the full list, along with descriptions of the work of all this year’s winners, please visit technologyreview.com/ supertopic/2023-innovators starting September 12. For instance, at AI Fund, a venture studio that I lead, I’ve been privileged to participate in projects that apply AI to maritime shipping, relationship coaching, talent management, education, and other areas. Because many AI technologies are new, their application to most domains has not yet been explored. In this way, knowing how to take advantage of AI gives you numerous opportunities to collaborate with others. Looking ahead, a few developments are especially exciting. Prompting: While ChatGPT has popularized the ability to prompt an AI model to write, say, an email or a poem, software developers are just beginning to understand that prompting enables them to build in minutes the types of powerful AI applications that used to take months. A massive wave of AI applications will be built this way. 7/28/23 5:01 PM 35 Innovators Under 35 75 I spent a long time trying to get aircraft to fly autonomously in formation to save fuel (similar to birds that fly in a V formation). In hindsight, I executed poorly and should have worked with much larger aircraft. I tried to get a robot arm to unload dishwashers that held dishes of all different shapes and sizes. In hindsight, I was much too early. Deep-learning algorithms for perception and control weren’t good enough at the time. About 15 years ago, I thought that unsupervised learning (that is, enabling machine-learning models to learn from unlabeled data) was a promising approach. I mistimed this idea as well. It’s finally working, though, as the availability of data and computational power has grown. NICO ORTEGA Vision transformers: Text transformers— language models based on the transformer neural network architecture, which was invented in 2017 by Google Brain and collaborators—have revolutionized writing. Vision transformers, which adapt transformers to computer vision tasks such as recognizing objects in images, were introduced in 2020 and quickly gained widespread attention. The buzz around vision transformers in the technical community today reminds me of the buzz around text transformers a couple of years before ChatGPT. A similar revolution is coming to image processing. Visual prompting, in which the prompt is an image rather than a string of text, will be part of this change. AI applications: The press has given a lot of attention to AI’s hardware and software infrastructure and developer tools. But this emerging AI infrastructure won’t succeed unless even more SO23-back_35.indd 75 valuable AI businesses are built on top of it. So even though a lot of media attention is on the AI infrastructure layer, there will be even more growth in the AI application layer. These areas offer rich opportunities for innovators. Moreover, many of them are within reach of broadly tech-savvy people, not just people already in AI. Online courses, open-source software, software as a service, and online research papers give everyone tools to learn and start innovating. But even if these technologies aren’t yet within your grasp, many other paths to innovation are wide open. Be optimistic, but dare to fail That said, a lot of ideas that initially seem promising turn out to be duds. Duds are unavoidable if you take innovation seriously. Here are some projects of mine that you probably haven’t heard of, because they were duds: It was painful when these projects didn’t succeed, but the lessons I learned turned out to be instrumental for other projects that fared better. Through my failed attempt at V-shape flying, I learned to plan projects much better and frontload risks. The effort to unload dishwashers failed, but it led my team to build the Robot Operating System (ROS), which became a popular open-source framework that’s now in robots from self-driving cars to mechanical dogs. Even though my initial focus on unsupervised learning was a poor choice, the steps we took turned out to be critical in scaling up deep learning at Google Brain. Innovation has never been easy. When you do something new, there will be skeptics. In my younger days, I faced a lot of skepticism when starting most of the projects that ultimately proved to be successful. But this is not to say the skeptics are always wrong. I faced skepticism for most of the unsuccessful projects as well. As I became more experienced, I found that more and more people would agree with whatever I said, and that was even more worrying. I had to actively seek out people who would challenge me and tell me the truth. Luckily, these days I am surrounded by people who will tell me when they think I’m doing something dumb! 7/28/23 5:01 PM 76 35 Innovators Under 35 While skepticism is healthy and even necessary, society has a deep interest in the fruits of innovation. And that is a good reason to approach innovation with optimism. I’d rather side with the optimist who wants to give it a shot and might fail than the pessimist who doubts what’s possible. Take responsibility for your work As we focus on AI as a driver of valuable innovation throughout society, social responsibility is more important than ever. People both inside and outside the field see a wide range of possible harms AI may cause. These include both short-term issues, such as bias and harmful applications of the technology, and long-term risks, such as concentration of power and potentially catastrophic applications. It’s important to have open and intellectually rigorous conversations about them. In that way, we can come to an agreement on what the real risks are and how to reduce them. Over the past millennium, successive waves of innovation have reduced infant mortality, improved nutrition, boosted literacy, raised standards of living worldwide, and fostered civil rights including protections for women, minorities, and other marginalized groups. Yet innovations have also contributed to climate change, spurred rising inequality, polarized society, and increased loneliness. Clearly, the benefits of innovation come with risks, and we have not always managed them wisely. AI is the next wave, and we have an obligation to learn lessons from the past to maximize future benefits for everyone and minimize harm. This will require commitment from both individuals and society at large. At the social level, governments are moving to regulate AI. To some innovators, regulation may look like an unnecessary restraint on progress. I see it differently. Regulation helps us avoid mistakes and enables new benefits as we move into an uncertain future. I welcome regulation that calls for more transparency into the opaque workings of large tech companies; this will help us understand their impact and steer them toward achieving broader societal benefits. Moreover, new SO23-back_35.indd 76 regulations are needed because many existing ones were written for a pre-AI world. The new regulations should specify the outcomes we want in important areas like health care and finance—and those we do not want. But avoiding harm shouldn’t be just a priority for society. It also needs to be a priority for each innovator. As technologists, we have a responsibility to understand the implications of our research and innovate in ways that are beneficial. Traditionally, many technologists adopted the attitude that the shape technology takes is inevitable and there’s nothing we can do about it, so we might as well innovate freely. But we know that’s not true. When innovators choose to work on differential privacy (which allows AI to learn from data without exposing personally identifying information), they make a powerful statement that privacy matters. That statement helps shape the social norms adopted by public and private institutions. Conversely, when innovators create Web3 cryptographic protocols to launder money, that too creates a powerful statement—in my view, a harmful one—that governments should not be able to trace how funds are transferred and spent. If you see something unethical being done, I hope you’ll raise it with your colleagues and supervisors and engage them in constructive conversations. And if you are asked to work on something that you don’t think helps humanity, I hope you’ll actively work to put a stop to it. If you are unable to do so, then consider walking away. At AI Fund, I have killed projects that I assessed to be financially sound but ethically unsound. I urge you to do the same. Now, go forth and innovate! If you’re already in the innovation game, keep at it. There’s no telling what great accomplishment lies in your future. If your ideas are in the daydream stage, share them with others and get help to shape them into something practical and successful. Start executing, and find ways to use the power of innovation for good. Andrew Ng is a renowned global AI innovator. He leads AI Fund, DeepLearning.AI, and Landing AI. This year we’re introducing a new feature to the 35 Innovators Under 35 competition. We’re naming an Innovator of the Year— someone whose work not only is exemplary but also manages to somehow capture the zeitgeist. For 2023 we’re happy to announce Sharon Li as our Innovator of the Year. Li received the highest overall numerical score from our judges, and her research on developing safer AI models is directly aimed at one of the most crucial and perplexing problems of our time. As AI models are released into the wild, this innovator wants to make sure they’re safe Sharon Li’s research could prevent AI models from failing catastrophically when they encounter unfamiliar scenarios. By Melissa Heikkilä 7/31/23 1:23 PM 35 Innovators Under 35 SARA STATHAS A s we launch AI systems from the lab into the real world, we need to be prepared for these systems to break in surprising and catastrophic ways. It’s already happening. Last year, for example, a chess-playing robot arm in Moscow fractured the finger of a seven-year-old boy. The robot grabbed the boy’s finger as he was moving a chess piece and let go only after nearby adults managed to pry open its claws. This did not happen because the robot was programmed to do harm. It was because the robot was overly confident that the boy’s finger was a chess piece. The incident is a classic example of something Sharon Li, 32, wants to prevent. Li, an assistant professor at the University of Wisconsin, Madison, is a pioneer in an AI safety feature called out-of-distribution (OOD) detection. SO23-back_35.indd 77 77 “know” is the weakness behind many AI disasters. Li’s work calls on the AI community to rethink its approach to training. “A lot of the classic approaches that have been in place over the last 50 years are actually safety unaware,” she says. Her approach embraces uncertainty by using machine learning to detect unknown data out in the world and design AI models to adjust to it on the fly. Outof-distribution detection could help prevent accidents when autonomous cars run into unfamiliar objects on the road, or make medical AI systems more useful in finding a new disease. “In all those situations, what we really need [is a] safety-aware machine-learning model that’s able to identify what it doesn’t know,” says Li. This approach could also aid today’s buzziest AI technology, large language models such as GPT-4 and chatbots such as ChatGPT. These models are often confident liars, presenting falsehoods as facts. This is where OOD detection could help. Say a person asks a chatbot a question it doesn’t have an answer to This feature, she says, helps AI models in its training data. Instead of making determine when they should abstain something up, an AI model using OOD from action if faced with something they detection would decline to answer. weren’t trained on. Li’s research tackles one of the most Li developed one of the first algo- fundamental questions in machine learning, rithms on out-of-distribution detection says John Hopcroft, a professor at Cornell for deep neural networks. Google has University, who was her PhD advisor. since set up a dedicated team to inteHer work has also seen a surge of intergrate OOD detection into its products. est from other researchers. “What she is Last year, Li’s theoretical analysis doing is getting other researchers of OOD detection was chosen to work,” says Hopcroft, who from over 10,000 submissions adds that she’s “basically creINNOVATOR ated one of the subfields” of as an outstanding paper by OF AI safety research. NeurIPS, one of the most presTHE YEAR tigious AI conferences. Now, Li is seeking a deeper We’re currently in an AI gold understanding of the safety risks rush, and tech companies are racrelated to large AI models, which ing to release their AI models. But most are powering all kinds of new online of today’s models are trained to identify applications and products. She hopes specific things and often fail when they that by making the models underlying encounter the unfamiliar scenarios typical these products safer, we’ll be better able of the messy, unpredictable real world. to mitigate AI’s risks. “The ultimate goal is to ensure trustTheir inability to reliably understand what they “know” and what they don’t worthy, safe machine learning,” she says. 7/28/23 5:01 PM 78 How culture drives foul play on the internet, and how we might protect ourselves. By Rebecca Ackermann Illustration by George Wylesol Fancy Bear Goes Phishing: The Dark History of the Information Age, in Five Extraordinary Hacks by Scott J. Shapiro FARRAR, STRAUS AND GIROUX, 2023 Online fraud, hacks, and scams, oh my T he world of online misdeeds is an eerie biome, crawling with Bored Apes, Fancy Bears, Shiba Inu coins, self-replicating viruses, and whales. But the behavior driving fraud, hacks, and scams on the internet has always been familiar and very human. New technologies change little about the fact that illegal operations exist because some people are willing to act illegally and others fall for the stories they tell. To wit: Crypto speculation looks a lot like online sports betting, which looks like offline sports betting; cyber hacking resembles classic espionage; spear phishers recall flesh-and-blood con artists. The perpetrators of these crimes lure victims with well-worn appeals to faith and promises of financial reward. In Fancy Bear Goes Phishing, Yale law professor Scott Shapiro argues that technological solutions can’t solve the problem because they can’t force people to play nice online. The best ways to protect ourselves from online tricks are SO23-back_books.indd 78 social—public policies, legal and business incentives, and cultural shifts. Shapiro’s book arrives just in time for the last gasp of the latest crypto wave, as major players find themselves trapped in the nets of human institutions. In early June, the US Securities and Exchange Commission went after Binance and Coinbase, the two largest cryptocurrency exchanges in the world, a few months after charging the infamous Sam Bankman-Fried, founder of the massive crypto exchange FTX, with fraud. While Shapiro mentions crypto only as the main means of payment in online crime, the industry’s wild ride through finance and culture deserves its own hefty chapter in the narrative of internet fraud. It may be too early for deep analysis, but we do have first-person perspectives on crypto from actor Ben McKenzie (former star of the teen drama The O.C.) and streetwear designer and influencer Bobby Hundreds, the authors of—respectively— Easy Money and NFTs Are a Scam/NFTs Are the Future. (More heavily reported Easy Money: Cryptocurrency, Casino Capitalism, and the Golden Age of Fraud by Ben McKenzie ABRAMS, 2023 NFTs Are a Scam/NFTs Are the Future: The Early Years: 2020–2023 by Bobby Hundreds MCD, 2023 books on the crypto era from tech reporter Zeke Faux and Big Short author Michael Lewis are in the works.) McKenzie testified at the Senate Banking Committee’s hearing on FTX that he believes the cryptocurrency industry “represents the largest Ponzi scheme in history,” and Easy Money traces his own journey from bored pandemic dabbler to committed crypto critic alongside the industry’s rise and fall. Hundreds also writes a chronological account of his time in crypto—specifically in nonfungible tokens, or NFTs, digital representational objects that he has bought, sold, and “dropped” on his own and through The Hundreds, a “community-based streetwear brand and media company.” For Hundreds, NFTs have value as cultural artifacts, and he’s not convinced that their time should be over (although he acknowledges that between 2019 and the writing of his book, more than $100 million worth of NFTs have been stolen, mostly through phishing scams). “Whether or not NFTs are a scam poses a philosophical question that 7/31/23 11:00 AM 79 SO23-back_books.indd 79 7/28/23 12:55 PM 80 wanders into moral judgments and cultural practices around free enterprise, mercantilism, and materialism,” he writes. For all their differences (a lawyer, an actor, and a designer walk into a bar …), Shapiro, McKenzie, and Hundreds all explore characters, motivations, and social dynamics much more than they do technical innovations. Online crime is a human story, these books collectively argue, and explanations of why it happens, why it works, and how we can stay safe are human too. To articulate how internet crime comes to be, Shapiro offers a new paradigm for the relationship between humanity and technology. He relabels technical computer code “downcode” and calls everything human surrounding and driving it “upcode.” From “the inner operations of the human brain” to “the outer social, political, and institutional forces that define the world,” upcode is the teeming ecosystem of humans and human systems behind the curtain of technology. Shapiro argues that upcode is responsible for all of technology’s impacts—positive and negative—and downcode is only its product. Technical tools like the blockchain, firewalls, or two-factor authentication may be implemented as efforts to ensure safety online, but they cannot address the root causes upstream. For any technologist or crypto enthusiast who believes computer code to be law and sees human error as an annoying hiccup, this idea may be disconcerting. But crime begins and ends with humans, Shapiro argues, so upcode is where we must focus both our blame for the problem and our efforts to improve online safety. McKenzie and Hundreds deal with crypto and NFTS almost entirely at the upcode level: neither has training in computer science, and both examine the industry through personal lenses. For McKenzie, it’s the financial realm, where friends encouraged him to invest in tokens to compensate for being out of work during the pandemic. For Hundreds, it’s the art world, which has historically been inaccessible to most and inhospitable for many—and SO23-back_books.indd 80 is what led him to gravitate toward streetwear as a creative outlet in the first place. Hundreds saw NFTs as a signal of a larger positive shift toward Web3, a nebulous vision of a more democratized form of the internet where creative individuals could get paid for their work and build communities of fans and artists without relying on tech companies. The appeal of Web3 and NFTs is based in cultural and economic realities; likewise, online scams happen because buggy upcode—like social injustice, runaway capitalism, and corporate monopolies—creates the conditions. Edward Snowden, who leaked classified information from the US National Security Agency in 2013, cross legal boundaries for what they believe to be expressly moral reasons. Bitcoin, meanwhile, may be a frequent agent of crime but was in fact created to offer a “trustless” way to avoid relying on banks after the housing crisis and government bailouts of the 2000s left many wondering if traditional financial institutions could be trusted with consumer interests. The definition of crime is also upcode, shaped by social contracts as well as legal ones. “If we are committing serious crimes like fraud, it is crucially important that we find ways to justify our behavior to others, and crucially, to ourselves.” Constructing downcode guardrails to allow in only “good” intentions won’t solve online crime because bad acts are not so easily dismissed as the work of bad actors. The people who perpetrate scams, fraud, and hacks—or even participate in the systems around it, like speculative markets— often subscribe to a moral rubric as they act illegally. In Fancy Bear, Shapiro cites the seminal research of Sarah Gordon, the first to investigate the psychology of people who wrote computer viruses when this malware first popped up in the 1990s. Of the 64 respondents to her global survey, all but one had developmentally appropriate moral reasoning based on ethics, according to a framework created by the psychologist Lawrence Kohlberg: that is, these virus writers made decisions based on a sense of right and wrong. More recent research from Alice Hutchings, the director of the University of Cambridge’s Cybercrime Centre, also found hackers as a group to be “moral agents, possessing a sense of justice, purpose, and identity.” Many hackers find community in their work; others, like In NFTs Are a Scam/NFTs Are the Future, Hundreds interviews the renowned tech investor and public speaker Gary Vaynerchuk, or “Gary Vee,” a figure he calls the “face of NFTs.” It was Vee’s “zeal and belief” that convinced Hundreds to create his own NFT collection, Adam Bomb Squad. Vee tells Hundreds that critics “may be right” when they call NFTs a scam. But while some projects may be opportunistic rackets, he hopes the work he makes is the variety that endures. Vee might be lying here, but at face value, he professes a belief in a greater good that he and everyone he recruits (including the thousands of attendees at his NFT convention) can help build—even if there’s harm along the way. McKenzie spends much of two chapters in Easy Money describing his personal encounters with FTX’s Bankman-Fried, who was widely called the “King of Crypto” before his fall. Bankman-Fried professes to believe in crypto’s positive potential; indeed, he has claimed on the record many times that he wanted to do good with his work, despite knowing at points that it was 7/31/23 11:00 AM But wait, there’s more. Lots more. You’re already a subscriber. Activate your account and start enjoying: • Unlimited web access • Exclusive digital stories • Access to 120+ years of publication archives technologyreview.com/subonly houseads.suite.indd 11 9/7/22 4:00 PM 82 potentially fraudulent. McKenzie struggles to understand this point of view. “If we are committing serious crimes like fraud,” he speculates, “it is crucially important that we find ways to justify our behavior to others and crucially, to ourselves.” While this rationalization certainly doesn’t excuse any crimes, it explains how people can perpetrate eye-boggling fraud again and again, even inventing new ways to scam. The human upcode that makes each of us see ourselves as the protagonist of our story is powerful, even and maybe especially when billions of dollars are at stake. Despite his research, McKenzie did gamble on crypto—he shorted tokens on a specific, and incorrect, timeline. He doesn’t disclose how much he lost, but it was an amount that “provokes an uncomfortable conversation with your spouse.” He’s hardly the only savvy individual in history to fall for a risky pitch; our brains make it painfully easy to get scammed, another reason why solutions that rely entirely on computer code don’t work. “The human mind is riddled with upcode that causes us to make biased predictions and irrational choices,” Shapiro writes. Take the “representativeness heuristic,” which leads us to judge something by how much it resembles an existing mental image—even if that may lead us to overlook crucial information. If an animal looks like a duck and quacks like a duck, the representativeness heuristic tells us it can swim. Phishing scams rely on this rush to pattern matching. For example, Fancy Bear, the titular Russian hacking group of Shapiro’s book, used a visually and tonally convincing message to attempt to hack into Hillary Clinton campaign staffers’ email accounts in 2016. It worked. Also coming into play for scams, fraud, and hacks are the “availability heuristic,” which leads us to remember sensational events regardless of their frequency, and the “affect heuristic,” which leads us to emphasize our feelings about a decision over the facts, inflating “our expectations about outcomes we like”—such as winning a huge payout on a gamble. When Hundreds was concerned about whether NFTs were a good investment, he reached SO23-back_books.indd 82 out to a friend whose belief was steadfast and found himself calmed. “It was that sense of conviction that separated the losers from the winners,” he writes, even when the facts might have supported stepping back. The marketing pitch of communal faith and reward, the enticement to join a winning team, feeds a human social instinct—especially as more offline modes of connection are faltering. It’s telling that after the SEC brought charges against Coinbase, the company responded by issuing a pro-crypto NFT, imploring its community to offer support for the struggling of the 2016 Clinton email hack, the billions lost by investors in the volatile crypto industry, and billions more lost through crypto hacks and scams. Shapiro argues that the efforts of the antivirus and antihacking industry to code guardrails into our online systems have failed. Fraud goes on. Instead, we must reexamine the upcode that has fostered and supported online crimes: “our settled moral and political convictions on what we owe one another and how we should respect security and privacy.” For Shapiro, effectively addressing online fraud, hacks, and scams requires Technological innovation does not change our fundamental behavior as humans, but technology has brought speed and spread to the gambling table. A single perpetrator can reach more victims faster now that the global world is connected. industry by minting it. (Coinbase and the minting platform Zora promise to donate the mint fees they’ll receive from consumers to pro-crypto advocacy.) The crypto industry rose to power on this kind of faith-based relationship, and it continues to appeal to some: more than 135,000 of the Coinbase tokens have been minted since the SEC suit was announced. Beyond money, “we’re just as motivated by identity and community (or its upside-down cousin, tribalism),” writes Hundreds, “and the most fervent contemporary movements and trends masterfully meld them all together. The only thing that feels as good as getting rich is doing so by rallying around an impassioned cause with a band of like-minded friends.” Technological innovation does not change our fundamental behavior as humans, but technology has brought speed and spread to the gambling table. A single perpetrator can reach more victims faster now that the global world is connected. The risks are higher now, as clearly demonstrated by the headline-exploding results political, economic, and social shifts such as creating incentives for businesses to protect customers and penalties for data breaches, supporting potential hackers in finding community outside of crime, and developing government and legal policies to prevent illicit payment through mechanisms like cryptocurrencies. Shapiro admits that shifting upcode this way will likely take generations, but the work has already started. The SEC’s recent moves against crypto exchanges are promising steps, as are the FTC’s public warnings against scammy AI claims and generative AI fraud. Growing public awareness about the importance of data privacy and security will help too. But while some humans are working on evolving our social systems, others will continue to hunt online for other people’s money. In our lifetimes, fraud, hacks, and scams will likely always find a home on the internet. But being aware of the upcode all around us may help us find safer paths through the online jungle. Rebecca Ackermann is a writer and artist in San Francisco. 7/28/23 12:55 PM Help shape the future of tech. Take part in original research and gain valuable business insights into today’s most important technology trends. Join the Global Insights Panel. • Participate in research programs and surveys (each takes about 5 to 10 minutes to complete). • Receive the latest tech news and updates in our exclusive newsletters. • Share experiences, weigh in on ideas from Insights, and form collaborations via our Global Insights Panel LinkedIn group. • Receive special promotions and discounts for MIT Technology Review conferences and subscriptions. Scan this code or visit us at the link below to join the Global Insights Panel today for free: technologyreview.com/GlobalPanel InsightsGP.indd 1 6/1/23 9:13 AM 84 Field notes Right: View of Godalming, Surrey, UK, population 21,000. Wasted heat from computers is transformed into free hot water for homes. By Luigi Avantaggiato Using heat generated by computers to provide free hot water was an idea born not in a high-tech laboratory, but in a battered country workshop deep in the woods of Godalming, England. “The idea of using the wasted heat of computing to do something else has been hovering in the air for some time,” explains Chris Jordan, a 48-year-old physicist, “but only now does technology allow us to do it adequately. “This is where I prototyped the thermal conductor that carries heat from computer processors to the cylinder filled with water,” he says, opening his workshop door to reveal a 90-liter electric boiler. “We ran the first tests, and we understood that it could work.” Jordan is cofounder and chief technology officer of Heata, an English startup that has created an innovative cloud network where computers are attached to the boilers in people’s homes. Next to the boiler is a computer tagged with a sticker that reads: “This powerful computer server is transferring the heat from its processing into the water in your cylinder.” A green LED light indicates that the boiler is running, Jordan explains. “The machine receives the data and processes it. Thus we are able to transfer the equivalent of 4.8 kilowatt-hours of hot water, about the daily amount used by an average family.” When you sign up with Heata, it places a server in your home, where it connects SO23-back_notes.indd 84 Servers that work from home via your Wi-Fi network to similar servers in other homes—all of which process data from companies that pay it for cloud computing services. Each server prevents one ton of carbon dioxide equivalent per year from being emitted and saves homeowners an average of £250 on hot water annually, a considerable discount in a region where 13% of the inhabitants struggle to afford heat. The Heata trial, funded by a grant from Innovate UK, a national government agency, has been active in Surrey County for more than a year. To date, 80 units have been installed, and another 30 are slated to have a boiler to heat by the end of October. Heata’s solution is “particularly elegant,” says Mike Pitts, deputy challenge director of Innovate UK, calling it a way to “use electricity twice—providing services to a rapidly growing industry (cloud computing) and providing domestic hot water.” The startup is now part of Innovate UK’s Net Zero Cohort, having been identified as a key part of the push to achieve an economy where carbon emissions are either eliminated or balanced out by other technologies. Heata’s process is simple yet introduces a radical shift toward sustainable management of data centers: instead of being cooled with fans, which is expensive and energy intensive, computers are cooled by a patented thermal bridge that transports the heat from the processors toward the shell of the boiler. And rather than operating with a data center located in an energy-intensive location, Heata works as an intermediary for computing: it receives workloads and distributes them to local homes for processing. Businesses that need to process data are using the Heata network as a sustainable alternative to traditional computing. The company has created what Heata’s designer and cofounder Mike Paisley describes as a diffuse data center. Rather than cooling a building that holds many servers, he explains, “our model of sustainability moves data processing [to] where there is need for heat, exploiting thermal energy waste to provide free hot water to those who need it, transforming a calculation problem into a social and climatic advantage.” The people involved in the Heata experiment are diverse in age and household composition, and their reasons for participating are varied: a need to save on bills, a love for the environment, an interest in helping combat climate change, and fascination with seeing a computer heat the water. Among the satisfied customers is Helen Whitcroft, mayor of Surrey Heath. “We started reducing our carbon footprint many years ago by installing photovoltaic panels,” she says. “We recently bought batteries to store the energy we produce. Curiosity also moved us: it didn’t seem possible that a computer could heat water, but it works.” Luigi Avantaggiato is an Italian documentary photographer. 8/3/23 8:44 AM 85 Field notes Left: Flats in Godalming, Surrey, UK. Over 4 million people in the UK struggle to afford heat. SO23-back_notes.indd 85 Below: The Heata team among the trees at Wood Farm, Godalming, where the idea originated. 8/2/23 6:45 PM 86 Field notes Above: A laser cutter produces insulation for the Heata unit, which harnesses excess heat from cloud computing. Left: Andrew, a mechanical engineer, installs the Heata unit in an apartment in Surrey. At 75% utilization, the Heata unit will provide around 80% of an average UK household’s hot water. Below: Parts of the Heata unit before assembly. Homeowner James Heather on his Heata: “We no longer need the energy for cooling the compute units, and we don’t need the energy for heating our hot water either, because we’re using the waste heat from the unit to do it.” SO23-back_notes.indd 86 8/2/23 6:45 PM Field notes 87 A batch of heat pipes at Heata Labs. Heata’s CTO, Chris Jordan, in his workshop. SO23-back_notes.indd 87 Dave, a radio engineer, tests the operation of the server at Heata Labs. 8/2/23 6:45 PM 88 A cell that does it all For 25 years, embryonic stem cells have been promising and controversial in equal measure. How far have they really come? From “The Troubled Hunt for the Ultimate Cell” (1998), by Antonio Regalado: “If awards were given for the most intriguing, controversial, underfunded and hush-hush of scientific pursuits, the search for the human embryonic stem (ES) cell would likely sweep the categories. It’s a hunt for the tabula rasa of human cells—a cell that has the potential to give rise to any of the myriad of cell types found in the body. If this mysterious creature could be captured and grown in the lab, it might change the face of medicine, promising, among other remarkable options, the ability to grow replacement human tissue at will … [but] these cells are found only in embryos or very immature fetuses, and pro-life forces have targeted the researchers who are hunting for ES cells, hoping to stop their science cold. In addition, the federal government has barred federal dollars for human embryo research, pushing it out of the mainstream of developmental biology. To make matters worse, human ES cells could conceivably provide a vehicle for the genetic engineering of people, and the ethical dilemmas surrounding human cloning threaten to spill over onto this field.” Update from the author (2023): The debate lasted years, but science prevailed over religion in the stem-cell wars of the early 2000s. Now research on ES cells is paid for by the US government. Yet biology keeps offering surprises. The latest? Research shows stem cells in the lab can self-assemble back into “synthetic” embryos, shockingly similar to the real thing. And that’s the next debate. MIT Technology Review (ISSN 1099-274X), September/October 2023 issue, Reg. US Patent Office, is published bimonthly by MIT Technology Review, 196 Broadway, 3rd floor, Cambridge, MA 02139. Entire contents ©2023. The editors seek diverse views, and authors’ opinions do not represent the official policies of their institutions or those of MIT. Periodicals postage paid at Boston, MA, and additional mailing offices. Postmaster: Send address changes to MIT Technology Review, Subscriber Services, MIT Technology Review, PO Box 1518, Lincolnshire, IL. 60069, or via the internet at www.technologyreview.com/customerservice. Basic subscription rates: $120 per year within the United States; in all other countries, US$140. Publication Mail Agreement Number 40621028. Send undeliverable Canadian copies to PO Box 1051, Fort Erie, ON L2A 6C7. Printed in USA. Audited by the Alliance for Audited Media. SO23-back_archive.indd 88 7/29/23 8:25 AM A DV E R T I S E M E N T The Green Future Index 2023 The Green Future Index 2023 is the third edition of the comparative ranking of 76 nations and territories on their ability to develop a sustainable, low-carbon future. It measures the degree to which economies are pivoting toward clean energy, industry, agriculture, and society through investment in renewables, innovation, and green policy. The index ranks the “green” performance of countries and territories across five pillars: • Carbon emissions • Energy transition • Green society • Clean innovation • Climate policy Green leaders The greening middle Climate laggards Climate abstainers Countries that have gone up in the ranking since last year Countries that have retained the same ranking as last year Countries that have gone down in the ranking since last year While the index ranks 76 countries, this map only features a selection of the overall data. Overall top 10 Rank 2023 Rank 2022 Rank 2023 Rank 2022 1 1 Iceland . . . . . . . . . . . . . . . . . . . . . . . . . 6.69 6 3 Netherlands ................. 6.22 2 6 Finland . . . . . . . . . . . . . . . . . . . . . . . . . 6.68 7 4 United Kingdom .......... 6.12 3 5 Norway. . . . . . . . . . . . . . . . . . . . . . . . . 6.37 8 10 South Korea ................. 6.00 4 2 Denmark . . . . . . . . . . . . . . . . . . . . . . 6.34 9 7 France .......................... 5.99 5 9 Sweden . . . . . . . . . . . . . . . . . . . . . . . . 6.33 10 13 Spain ............................ 5.92 Territory Score/10 The Green Future Index 2023 was produced in association with Premier partner GFI23.single.final.indd 1 Gold partner Silver partner Territory Score/10 Scan the QR code to experience the interactive index, view the data, and download the full report or visit technologyreview.com/gfi Interested in partnering with us? Contact: insights@technologyreview.com 5/16/23 11:16 AM Give the gift of knowledge. Gifting an annual subscription to MIT Technology Review will keep the tech-obsessed people on your gift list inspired all year long. A gift subscription includes: • In-depth, thoughtful stories on AI, climate change, biotech, and more • Expert insights on emerging technologies impacting the world • Unlimited access to our website, app, and magazine archives • Discounts on all our signature events Scan this code to purchase a gift subscription or go to TechnologyReview.com/GiveAGift Untitled-1 1 8/3/23 13:10