>> Jonathan Grudin: Good afternoon. I'm happy to be able to welcome Tom
Finholt to give a talk here today. Tom is doing a sabbatical at the University of
Washington this semester and is over here collaborating, talking with people in
MSR while he's here. He is the senior associate dean at the School of
Information at Michigan and is an expert and has long done research in the areas of cyber infrastructure and collaboratories. He's been one of the key people in the NSF and NIH collaboratory oversight reviewing area.
But today he's going to be talking not so much about collaboratories and cyber infrastructure but about sustainability and social networks. And I know that there are a good number of people here who are also interested in both the issues of sustainability and social networks, but I haven't yet heard too much done bringing those two together, so I'm looking forward to this presentation.
Thank you, Tom.
>> Tom Finholt: Thank you, Jonathan. So I can just assume this mic is live now? Okay. Thanks.
Welcome, everybody. Good to see you, and certainly enjoying the Seattle weather today. We'll see how it goes in a couple weeks.
So the work that I'm going to talk with you about this afternoon is a relatively new line of research we've initiated at the University of Michigan, and we've just recently received a new NSF award to support this. So I don't have so much in the way of results, but I've got a lot in the way of hypotheses and direction that we're going to go.
And I want to start by sort of motivating the conversation -- I'm sure for this audience I don't have to motivate it too much -- but we are an economy and a society that is highly dependent on fossil fuel and carbon-based energy, oil be the principal driver of the economy. And the best case scenario is peak oil is out there 40 years from now and the worst case is peak oil happened a decade ago.
It's inescapable that these resources are not renewable, and so there's a very strong societal imperative to reduce our dependence on these kinds of fuels.
And as if that was not enough, the combustion of these fuels produces a secondary and very dramatic problem associated with carbon load and increased global warming and climate change and so forth.
So the way we approach this problem was to say what are the things that we can do in our everyday existence that may help with the reduction of our dependence on these non-renewable energy sources and that might contribute in some way to reducing this climate burden.
And I don't know if you've read the New York Times in the last two weeks. I think there was an editorial by a noted environmentalist who said these kind of
incremental individual-level behaviors are meaningless and the main thing is to pass cap and trade and create the giant macroeconomic levers and so forth.
But there was, as you might imagine, a rather Folsom response to that suggesting that you don't create a constituency or an advocacy for those kinds of broader scale changes without enabling people to make the kinds of changes in their individual lives that let them feel like they are contributing and so forth. And the we actually feel we're going to be able to go beyond that.
So there is a kind of a characterization of the weekend environmentalist who is, you know, I went to Lowe's and wherever and bought my carton of CFL bulbs and put them in and now I've done my bit for the environment and so forth, and we want to try to go a little bit beyond that.
And one of our organizing constructs is dematerialization. And it's not something from Buck Rubon's [phonetic] eye or something that you saw in Ghostbusters. It actually is an economic construct.
And what it refers to is the ability to increase gross domestic product. In other words, increase the well-being of individuals in the society without also increasing ad infinitum their dependence on material resources.
So I know some of us in this room are old enough to have been familiar with the literature on limits to growth that was popular in the late '60s and '70s, and what it did was it did this kind of Malthusian play where it took the consumption of materials, laid it out on an extrapolation and said, you know, my God, we're going to run out of everything in the next five to ten years and so forth.
Well, that hasn't happened. And one of the reasons why that hasn't happened is through the miracle of dematerialization. And let me just make a brief side trip here to explain what we're talking about.
So one kind of canonical example of dematerialization is in the domain of telecommunications. And I know that you know what that is from your past, right?
This is a bundle of copper telephone wires. Just to give you an idea of the capacity, we'll say something like 24 voice channels, something like 1.5 megabits per second and forth.
And that's what the whole world used to run on was copper wire. Well, copper is horrendously heavy, for one. It is increasingly scarce. It involves quite a bit nasty extractive industry to mine it and so forth.
And what's happened over the last two decades is copper has been largely replaced by fiberoptics. And this replacement is a nice example of dematerialization because this is just glass, and the capacity is dramatically increased. So something on the order of 32,000 voice channels, 2.5 gigabits per second and so forth.
So you can play that kind of game throughout the economy. And, in fact, economists have noted that dematerialization is paying off, particularly in the developing economies. If you took where we were back in Jimmy Carter's day, if you remember his presidential speeches, some would characterize them as doom and gloom talking about the need to modernize the energy infrastructure and so forth.
A lot of that actually was taken to heart. In the United States today, if we'd continued consuming on the path we were in in the '70s, we'd be probably 35 to
40 percent higher in terms of consumption.
So it's not like we're doing well. So we still have a huge dependence on foreign oil. We're still producing a vastly disproportionate share of the world's carbon outputs, but it could have been a lot worse. And the reason it's not a lot worse is because of this miracle of dematerialization.
And the developing economies have this down to a science -- or the developed economies have to down to a science. The developing economies, there's evidence that Brazil and China are now tipping over into this period of dematerialization and so forth.
So we took this dematerialization metaphor and played around with it a little bit in the context of some more mundane and everyday concerns that we have, and the principal one was that we received an award from the National Science
Foundation to convene a virtual school of computational science and engineering as a partner activity to the award of the Petascale facility at Illinois. And, no, that's not actually the Petascale facility. Notice it says interim system. If you've read the headlines recently, IBM just pulled out of the Petascale agreement, so there actually isn't a blue waters machine right now. But the money has been funded and you can go visit the building in Champaign --
>>: [inaudible]
>> Tom Finholt: Yeah, who knows. Well, it's interesting, the Blue Waters
Building is the top lead certification even though it's a giant, you know, machine room and so forth.
So they had this need to convene this virtual school, and they could have gone about it in the traditional way and reserved a giant lecture hall on the Illinois campus and flown in the dozens of eager beaver graduate students who are hot to learn how to parallelize their code and so forth, and then a stream of talking heads could parade in front of them and edify them and learning would happen.
What we proposed instead was a virtual model of this school, taking advantage of some experiments we'd been doing related to the work that Jonathan mentioned early with the collaboratories and exploiting capabilities of -- this is not compressed HD, but it is packetized HD, moving around within the big 10's gigapop infrastructure.
So we can exploit that to create a very vivid and, as you can see in this case, big brother-ish type experience. And there were multiples of these all over the place.
So what we were essentially replacing was convening everybody in a single place with using an advanced network and video conferencing structure to create a multi-site conference where there had been a single-site conference.
And that was, you know -- to be honest, the main motivation for doing this was because a lot of us were really geeked about this HD video technology and we wanted to try to demonstrate some of the capabilities to the leadership of these campuses and so forth.
Then we got to think about it a little bit, and especially since we'd seen some of these presentations on dematerialization, we realized that we had been doing dematerialization while convening this multi-site conference. And so we actually looked at who attended these things and so forth, and we did kind of a thought experiment.
And the one arm, if you consider everybody coming to this single location in
Champaign, here's where all the participants were, and these are their flights to get to champagne. Those of you who have been to champagne know that you can't actually get there directly. You have to go through O'Hare first and then take a little hop down the state. But we have magic airplanes for purposes of this analysis.
So this is what it would have looked like. And in particular, it's something like
87,000-person -- or passenger miles, and that sort of an imputed energy load of about 79,000 kilowatt hours. If we had done nothing different, that's what would have happened at that meeting.
As an artifact of doing this multi-site version -- and you can see, the multi-sites weren't exactly optimized for efficient travel in some cases. There's some people coming from parts of the east and going to some of these southern institutions.
For some reason this content doesn't appeal to anybody west of the Mississippi with one exception. Maybe they go to meetings in San Diego for their --
>>: [inaudible]
>> Tom Finholt: Yeah, who knows what's going on there.
And here you can see a fairly dramatic reduction in the passenger miles and a fairly substantial reduction in the energy footprint. And this is through one relatively modest intervention.
And so the way we started thinking about this was we realized we were characterizing the power draw of the meeting and that it became possible now to start thinking of meetings as another kind of appliance or device that we might put on the grid and that there were things that we could do to change the nature of those meetings to dramatically alter the energy consequence. And in this case this savings of about 30,000 passenger miles, about 27,000 kilowatts, and that's
depending on whose calculations you believe, somewhere on the order of 15 to
19 tons of carbon dioxide. And the principal savings is by eliminating the need for the airplane travel.
So I'm not going to talk about this here, but there was an analysis done by actually a climate scientist who was attending an international summit of climate scientists and was struck by the irony of a community of climate scientists all hopping into jet airplanes to fly to London to have their -- or Copenhagen as it was last year -- to have their meeting about the threat of global climate change.
Jet travel is pretty close to the worst thing that we can do, and yet it's ubiquitous.
It's our ubiquitous solution to all forms of networking and convening and so forth.
So we were struck by this issue of estimating the power draw of meetings, and then we started thinking about whether we might be able to do the same thing for individuals. And it turns out that there is an interesting application on the web that was developed by Saul Griffith [phonetic] called WattzOn -- W-a-t-t-z On -- and it lets you input numbers, depending on your patience, you know, 10 or 15, or you can get as detailed as you want, and essentially lets you figure out what you -- think of yourself as a light bulb. What is your power draw?
So I went through this exercise myself. There's Saul right there. This is actually the alpha version of the website. They've now productized it and it's focused much more on home energy consumption and less on overall energy consumption, but this site is still running. I encourage you to go around and play with it.
So I input my numbers into this thing. And it turns out that if I were a light bulb, I would be rated at 4.8 kilowatts. So what constitutes that draw for me? Well, transportation is a fairly significant component of it. And actually since I became an associate dean, that component diminished. But when I was a regular faculty member it was twice as big as that. And, in fact, I've seen people who have on the order of 70 to 80 jet airplane trips a year, and you can imagine that that just blows it up.
If you remember from the previous slide about two kilowatts is on the order of the average consumer. And a lot of academics I put in are up above 11, 12, 15, 22.
And so we are -- not only are we this stinking ninth, we're also polluting tenth or something like that.
Shelter is the other significant component there. So I did these as if I was living in Ann Arbor. This component would be smaller in Seattle because your heating season is shorter and you don't have a cooling season.
And then the last category is stuff. Now, stuff is actually an important consideration here because while these things have to do with energy burned or consumed for me to do something, like to get from Seattle to Pittsburgh or to keep my family warm during the darkest days of the Michigan winter, this is like my iPhone.
So if we think about the iPhone, it's consuming power, it's consuming infrastructure resources that have to be powered presumably in the network, the wireless network, the WiFi network and so forth, but it also embodies a certain energy cost because someone had to make it, it had to be fabricated, it had to be shipped to me, someone had to build a store, an AT&T store or an Apple store to self it to the me and so forth.
And so the interesting thing about this WattzOn site is it's now incorporating all that embody energy as well as your actual energy consumption, and that's an important thing to keep in mind as we try to figure out where are the dematerialization wins. They're not always so obvious.
And let me characterize one that is sort of popular but is totally bogus, which is the transition from pressing CDs to distribution of MP3s. So on the first glance you would look at that and say that must a huge win because we're not pressing all that plastic, we're not driving it around the country, we're not -- you know, all record stores, you name them, they're all gone as far as I can tell, and we're moving these pure bits around and so forth.
Well, not really, because that music has to be stored somewhere. Those servers have to be spun up and cooled. And the big problem is all the embody energy in the devices.
So if these things lasted as long as a washing machine does, it wouldn't be so horrendous. You know, washing machines people replace on the order of a decade or -- actually my dad was quite pleased to have gotten one to go 25 years.
These -- this one is probably at the end of its useful life. I think this is a 3G. I bought it probably two years ago. It was already obsolete when I bought it. The
4s in two weeks will be obsolete in, you know, onward, the thrust of progress and so forth.
And so what looks like a sort of immediate dematerialization win isn't always a win. And that suggests that you've got to drill down below the surface to understand what's really going on with a lot of these technologies if you want to do these kinds of energy accounting.
And the reason you have to do this is because if you want to try to benchmark or describe improvements, you've got to know where you're starting from. So you can sort of see our progression as we got into this by trying to characterize the draw of the meeting, and then we realized that people are the constituents or components of the meeting, so we want to drive down to that low level, and now we're stuck with this problem of we can't get thousands of subjects to sit down in front of Saul Griffith's website and answer this stuff thing. There's, I think, 500 and some things that have been characterized in terms of their embodied energy.
Only a super geek like myself and my colleague Eric Hofer and maybe 2- or
3,000 others would sit and do that.
So what we're looking for is a solution that characterizes this sort of energy footprint dynamically. So this is a relatively static slice. You know, you answer these questions once and then you can -- you can actually connect to your power bill now and it will update based on your electric and gas consumption.
If you connect through Fuelly it will actually keep track of your gasoline purchases for your car, and if you give the make and so forth it can impute mileage. But for the most part this is not a very attractive alternative for gathering these kinds of data.
So the reason that we want a solution that will be more tractable is that we've become interested in what we'd characterize as the energy costs of social networks.
Now, sometimes NSF program officers in particular are guilty of reading this and saying the energy costs of computer-based social networks. And we're actually about a much broader agenda, which is the cost of maintaining your social network over a host of infrastructures and technologies.
And the notable ones would be transportation, communications, and computer.
And I think reflectively we're oriented to talk about the computers and not think about the bus ride we took, the car we drove, you know, the commute and so forth.
So just to give you an idea of sort of the kinds of things that we're looking at, this is a visualization that was done by Aaron Koblin at UCLA. Do you know what it is?
>>: [inaudible]
>> Tom Finholt: No, it's not email.
>>: I can read it.
>> Tom Finholt: It's the transportation. It's airplanes. So these are the -- and what you've got over here is the clock running in eastern standard time. And if you watch it, you can see the ebb and flow of the departures and arrivals. So here the East Coast is waking up, it's activating. Now the West Coast wakes up.
Here you can see the incoming European flights over there and so forth.
So this air transport network is a fairly substantial component of maintaining social network. This is grandpa flying to visit the grand kids. This is families convening for weddings or funerals or anniversaries or reunions. Airplanes are a significant component of that maintenance and one that we think has not really been attended to.
So if you look at the general social survey and so forth, they're asking you how often did you communicate with someone, but they're not asking did you drive there, did you fly there, did you have a video conference. How did you maintain
that contact? And you need to know that if you're going to impute these sort of energy demands. So that's the air transportation network.
We're also concerned with automobiles and busses and other kinds of personal transportation. And then we're similarly concerned with the use of telephones and cellular grids and so forth. And then we're also interested in the use of social media.
So with respect to the transportation networks, we can pretty easily get people to self-report. Those are not -- except for a very small fraction of people who might be interested, those are fairly salient. Most of us can come up to on a reasonable approximation of how many plane trips we might take per annum you know, how many were short haul, medium haul, long haul and so forth.
It's much more difficult for us to say anything about how we're using computing or communication infrastructure to maintain Facebook status or to check in on foursquare or to send email and so forth.
And so this has led us to think carefully about how we might unobtrusively capture that behavior in a grounded way that lets us make accurate predictions or forecasts about energy consumption and then ultimately might help individuals and organizations make decisions about how they should allocate resources.
And for example, you know, jet travel might be very appropriate in the beginning of a collaboration, but as the collaboration gathers steam, maybe it needs to be maintained through some less material-intensive modality like video conferencing and so forth.
So just as we were starting this think about this problem, this movement was born around the quantified self. So are you familiar with this? Also known as self-tracking. They had a meeting just a few months ago. It was really kind of the first convening of the tribes. And I've got to tell you that I didn't go to the meeting, but my colleague, Paul Resnick, went and my colleague Eric Hofer has been to other QS meetings.
It is sort of a combination of camp revival, you know, maker fair and just out right eccentrics and weirdos because a good deal of it has been oriented around personal health. And some of the origins of this come from the performance athletics tradition. So if you are a championship cyclist or a champion runner, you are acutely sensitive to the state of your body, and you're very interested in tracking your performance. And that has to do with sort of inputs, you know, what did you eat, what was your training regime, what was the output in terms of did I gain efficiency. If I'm training, when do I need to tail down, when do I need to carbohydrate load and so forth. And basically that was available to a very small element of the population, you know, members of the Olympic team and so forth who had hoards of acolytes who stood around and kept track of that kind of stuff.
Well, what's happened with the revolution in personal electronics is that a lot of those capabilities are now accessible to ordinary people. And so you can buy systems that let you sort of track your activity levels, there's popular ones for
monitoring your sleep states, and as these things have grown and they've been associated with social media, you get the capacity to share that information with one another.
So there is this interesting opportunity now that you could realistically imagine that we could capture the behavior of individuals across a host of these infrastructures related to this support of social networks in a way that previously would have been intractable or very cumbersome. I mean, essentially you would have had to do some kind of exhaustive, you know, field surveys, phone interviewing and whatever, and those are notoriously prone to error. People's recollection decays pretty quickly. They use items on the instrument as a gauge for what they think they should be answering.
Have you ever heard of this phenomenon? You know, you ask people how much television you watch, and if you move the range around you can actually move their response around because they choose the middle -- they say, well, I must be an average TV watcher so I'll circle the middle. And so I can say 100 to 150 hours or 10 to 1,000 hours and they're right in there circling that middle.
So just as in many things, the gold standard would be the actual behavioral data.
And so this quantified self-movement, as crackpot as it occasionally can be, raises the possibility that there is an opening that we might imagine kind of a personal energy informatics that could then be aggregated and provide a data resource that could be exploited and try to tune some of these larger systems in ways that might be beneficial in terms of our larger goal of reducing dependence on these non-renewables and the carbon load.
So let me walk you through kind of quantified self 101. So this is one of the very popular gizmos in the QS world. It's a device called fitbit. Have you ever seen one before? Yeah.
So you essentially wear it, and it tracks something on the order of four our five to six parameters about you, mostly how many steps have you taken. So it's sort of like the old-fashioned pedometers except this thing can report on you. And you can upload these reports, and then it can actually tell you -- you can see down here your best all time number of steps per day and then up here -- in this case
Eric has set a weekly goal for steps, and are you getting close to them and so forth. And these could be aggregated across a population of users and you could start to get sort of a realtime diagnostic picture of the activity rates of your organization.
Now, I mean, set aside some of the Big Brother-ish aspects of this. You know, my colleagues like Paul Resnick have argued that this could be an important tool for motivating healthy behavior because now there's a leader board, and God knows Americans love nothing better than a leader board to show, you know, that I walked 18 percent more steps than you did, et cetera, et cetera.
So this is kind of the metaphor for this QS world is some fairly simple-minded device with the capacity to record on a narrow range of parameters and then they
can be uploaded to some location in the cloud and compared and processed and so forth.
And this is a similar kind of idea, but now in the domain of things that we're concerned with in this project. This is an effort on the Android platform by some colleague of ours in electrical engineering at Michigan. And what it is is a -- they it the powertutor, and it tells you not only what is your ask consumption on your device -- so, you know, sort of here you're seeing what the load is from the CPU and here's your networking demand and so forth -- but it actually breaks it out by application.
So you can start to see, oh, my God, you know, the amount of time I'm spending on Facebook, that is a significant drain on my battery. And so if I could change my behavior, then I would reduce it and I would be able to answer the phone when my girlfriend calls me and so forth.
So it's in kind of the microcosm. You have so sort of use your imagination to see where this kind of thing might go. But just as with this other QS stuff, you can imagine this kind of energy monitoring also expanding into the aggregate so that suddenly, as I have on the University of Michigan campus, 42,000-some students zipping around town, they are all now sort of sensors, if you will, telling me about the realtime consumption of resources within these infrastructures that are supporting their social networks and so forth. And so --
>>: What's the first step?
>> Tom Finholt: What?
>>: What's the first step up there, the screen shot?
>> Tom Finholt: This one? Oh, this one? It's an Android thing, so I don't know --
I mean, it's taking pictures of the screen. I don't know why that should be so power intensive. I mean, maybe they set their thumb down under or something like that.
>>: [inaudible] the time on each of these apps?
>> Tom Finholt: That I'm not so -- I mean, you can actually get that in another display. And so what you'd really like to know is -- I mean, there are some things that you do a lot but in a not very deep fashion. Like you might check in and look at Facebook or email or something. But then there's other things that you do infrequently that might be very intensive, like watching videos and if you have like that. So that would be another dimension that you would be interested in.
So an extension of this, then, is this growing market for energy-sniffing devices, I will call them. So this one is a thing called Kill A Watt, and you can see you can plug it in between your device and the wall and it will keep track of exactly how much juice you're burning by having all your stuff in a standby mode, which is scary large, actually. It's like having a trip in your faucet or something like that.
It's not very salient, but it turns out over the course of a year it's a lot of water.
Well, it's a lot of electricity that people are burning away. And you've heard of these -- anybody drive a Prius in here? So have you heard of these clubs of
Prius drivers who are -- you know, their high mileage or -- you know, they're trying to see if they can out do one another in efficiency, because you've got all those readouts on the -- yeah. So, I mean, there are people who are similarly oriented to these Kill A Watt devices and they're going around their houses, you know -- so we can say that that's kind of bizarre behavior.
And one of the things we've done in -- well, it's unusual. It's probably a little bit harsh to say it's bizarre. One of the things we've done in the NSF award is to say we want to target those kind of people as early adopters of personal energy technologies because they're going to go out and exercise them. And Larry
Smarr used to say that high performance computing was like a time machine because you could see in the moment what would become ubiquitous in the future.
And so in the same way we want to take these QS people and use them as our time travelers because we believe they're already ahead of the curve. They're already the ones wearing the can you have 24 hours a day, the little fitbit, plugging these things in all around their homes. They're an interesting breed of adopter because they're intrinsically interested in this self-tracking, and they will enthusiastically engage.
And so what we're trying to get from them is sort of a snapshot of this intensive use as a proxy for what these things might look like if they were deployed ubiquitously, and so instead of having 300 observations, if you had 300 million observations.
And the way to think about that is -- I think Barbosi [phonetic] did a study with cell data from Germany essentially being able to show sort of the passage of devices through the footprint of these things, and you're able to answer all kinds of questions about, you know, where people are going, what they're doing, even with these very simple parameters. We want to achieve the same kind of thing with these sort of electrical devices.
And then this one, watts up, actually will produce reports. Again, so you see the analogy to kind of fitbit-type technologies. These are now creating a market for data that might be generated by these things. And this is very early days.
And so one of the things that NSF is very excited about is whether we can engage people in the use of these kinds of technologies and then, correspondingly, what sorts of questions might we be able to address if we had that information.
And from our point of view, the main thing that we're concerned about initially is the tradeoff between the decision to use very energy intensive infrastructure to build and maintain these networks versus less intensive. And is there -- are there obvious sweet spots or wins that we can exploit theoretically to say that really isn't a good reason for you to take a plane trip. This would be an occasion
when a video conference would be completely adequate, but not a video conference like we are accustomed to. This is the quote I was looking for the other day. You know, the long, darkroom with the uninterested parties in the front and the angry person screaming at the end, which is sort of the H323 experience of video, enhanced video.
And as you can recall from that previous analysis of the multi-site meeting, we've got a pretty hefty power budget to deal with there just in a kind of ad hoc distribution. We gained back all of those kilowatts. It's actually something like on the order of 553 kilowatt power budget.
Well, there's no way that these technologies are chewing up that much juice. In fact, our back of the envelope estimate was that for the virtual school, it was probably something in the order of 50 kilowatts or something like that.
So you could do much more and still be ahead of the game in terms of the consumption on the jet airplane travel and so forth. And I think that's a challenge for the developers of these technologies to try to figure out what more would one do to make sure it more vivid, more realistic and so forth. And I think we've been in a period where we sort of vacillate between irrational exuberance about things like video conferencing and then these sort of thermidorian crashes where it all just sucks.
And I know Jonathan has been through at least three of those cycles and you've probably been through a couple of them yourself from the bell core [phonetic] days. And I think there's probably more that could be done. So that's one kind of question that we're interested in.
And then the other kind of question we're interested in is whether these large-scale infrastructures can be optimized and better balanced. And the metaphor here is the Green Light Project at UC San Diego.
So this is an effort that Tom Defanti and his colleagues have. They basically have a machine room that is completely self-contained, and so they know exactly what goes into it and what comes out of it. They know how much power goes in and how much heat and cooling and so forth have to go out.
And so you can set that thing up and then run simulations on it to try to see what the -- how you might balance and optimize the load. And we're thinking from the point of view of these larger infrastructures to support social networks, there may be things that you can do to sort of tune them in smart ways so that you build in capacity when it might be required. You know, from the transportation system point of view, there's clearly extra capacity built in around holiday periods to accommodate the extra travelers and that kind of thing. But there may be things that could be tuned on an even more micro basis within what I would characterize as microclimates of structure.
So, again, in Ann Arbor we have a unique microclimate, the 42,000 students, the
34,000-some staff, the 7,000-some faculty. These are all people who -- they are avid consumers of these technologies, and we think it's going to be very easy to
sort of get them to answer questions and put some of these devices on their systems and so forth.
>>: I have a question about measurement. So even in this simulation, you're measuring some parameters and not others. So I could decrease my energy drain by letting the air conditioning go up 20 degrees. But that might have consequences of that burning out machines. So when you talked about stuff, it was sort of interesting all the kind of embedded costs. How important is that in some of these simulations? So I can win the game locally by doing things that are not from somewhat workable perspective.
>> Tom Finholt: Yeah. Well, that's one of the things we're trying to address with this new award is to try to figure out where those wins are actually suboptimized.
And I think a lot of the rhetoric about machine rooms is exemplary of that kind of suboptimal approach.
I mean, they basically -- the machine room operator has in front of her this problem, you know, I've got to try to get this system to as close to equilibrium as I can. But you're absolutely right, they don't have to worry about whatever energy cost is embodied in the technology that they're looking at.
Now, the lead certification actually takes some of that into account. And I think one of the really interesting dematerialization arguments is that if you can figure out how to not eliminate but slow your facility's growth -- and this has to be an important concern even for Microsoft these days -- if you can diminish it, say if your facility's growth is on the order of 1 percent, which would not be an unreasonable number, if you can drop that below half a percent, the amortized savings are gigantic because now you don't have to cool that space you didn't build, you don't have to clean it, you don't have to furnish it and so forth.
And at the University of Michigan we've just completed a campaign where we were horrified to discover we actually have the largest physical campus of any institution in the world. We didn't know that when we started. And then we were horrified to discover that not only were we the biggest physical campus, we were growing it at the rate of 1.2 percent a year, which is truly not sustainable.
So we've now reduced that to under half a percent, which means, you know, tremendous benefits with respect to savings, money that can be put toward, you know, hiring faculty and all those other kinds of things that the university might actually be interested in doing.
But I think you are capturing the complexity of the problem. It's sort of like if you push here then it squeezes somewhere over here. And let me add just another layer of complexity, which is this principle known as Jovan's paradox.
So Jovan was, I'm going to say, an Oxford don or something like that. He was in one of those institutions. And I didn't include his picture here. He's a very proper
Victorian economist. His paradox -- and this was in the context of steam production. But the paradox was that as you make things more efficient, you can actually increase the demand because the price goes down.
So you can have this ironic consequence that we could make these things super efficient, but then we can actually offer the services, the cloud services, just to take an example out of thin air, and now people are going to see that lower price and they're going to come in.
And so even though we made them 50 percent more efficient, we increased the demand by 70 or 80 percent, and so that efficiency is moot in the face of the larger demand.
And it turns out that the energy economy has little pitfalls and traps like that all over the place, which is why you really need to take this sort of systematic perspective to try to understand the interdependencies across all of these different factors.
>>: Well, since -- yeah, so, I mean, they are all over the place. One example from a place I used to work was the part of the company that was involved with energy would say turn off your computers at night to save energy, but then the IT people would say turning them off and on every day, you know, leads to them burning out -- in the old days they used to say it leads to them burning out, so they wanted them on. And nobody had any answer as to what the global benefit was.
Similarly, though, you know, you mentioned the example of the CDs, the DVDs.
What about the Kill A Watt and the WattzUp devices? I mean, you get tens of thousands of people carrying those around, the production and recycling of those when their life is over, you know, what's the cost there?
One more example which is down the road for the transportation example, people who are -- people are very interested in that issue, technology and transportation, and that's Boeing, and they studied it for years, and their conclusion, which feeds right into this, is that you don't have to worry because the effects of the internet and the effects of networking is going to mean that people will start collaborations in more distant places and they're going to end up having to travel some time there.
Similarly, with social networking software, I can now keep in touch with lots of friends from high school and college much more efficiently, but just one or two times a decade I go and visit them, fly out to see somebody who I want have before they were out of sight --
>> Tom Finholt: You've trashed all the --
>>: Yeah. So, you're right, there are tremendous of these paradoxes, right?
>> Tom Finholt: Right. And then sort of figuring out from a systems point of view is one of the aspirations.
So the Kill A Watt and WattzOn, I think you have to look at those as kind of
Model Ts and to say that those are purposes-built devices for hobbyists to sort of
keep track of things and then -- you know, it's like ham radio operators, and they share stuff with one another and they can compete and that kind of thing.
The more notable transition is the design of the devices with those monitoring technologies, you know, built in. So it's already part of the device. It's another bunch of gates or something on one of the chips, and that's the thing that's monitoring do we spin up the fans or how do we allocate the juice across the thing.
And so now since it's already in the embodied budget for the thing, you might have paid a little bit extra for that, but then it's making the device that much more efficient, and therefore you can imagine payoffs down the road. But it can be opaque.
And particularly for consumers who avidly jump to the next thing, you know. In two weeks I guarantee there will be reports on CNN and the New York Times of the fan boys and girls lined up to get whatever new bauble Apple has produced, right?
And how many of them would be lining up if they actually had to pay the embodied cost of those devices. Or as people have proposed when you purchase an automobile, you also need to purchase the end-of-life plan, which I think you have to do with the Prius, don't you a little bit? Don't you have to pay ahead to cover the disposal cost of the battery? I think it's built into the --
>>: [inaudible]
>> Tom Finholt: Well, you didn't see it, but I think it's built into the cost of the device. And then I think Toyota subsidizes other aspects of it so it's not -- it's still painful, but --
>>: The batteries are very expensive, so that's probably why.
>> Tom Finholt: Right. And I think from a point of view of the life of the vehicle, that's the most problematic component.
So the last thing I would leave you with here is that in our studies of scientists and engineers collaborating -- and this is independent of whether they're thousands of miles away or tens of meters away, the thing they keep coming back to is I want some information about when do I switch from one modality to the other.
And it's almost like the old hedge used to be by IBM because nobody got fired for buying IBM. I suppose at one point people said the same thing about Microsoft, and perhaps it's Google today. I think it's the same thing about communication is nobody ever got fired for trying to have a face-to-face meeting or for organizing things in a face-to-face session.
So there's very little incentive to explore what the intersection might be between conventional meeting and technology and some of these last orthodox
alternatives that may be less intensive. And I think it's this inertia that we have to get over across all of our activities, and this is just one that's very salient to us.
We spent five years playing around with these incredible HD video technologies and we built these, of course, massively energy consuming tiled walls and had, you know, great teas with our pals on the other side of the continent and could never get anyone else interested in using those kinds of things.
And so what we hope to be able to show with some of this data is to start to address -- from the point of view of a surgeon in the Michigan medical school who right now sort of transparently hops into her car and putters around campus to go to meetings. If we were able to play back to her exactly how much of her life is lost in maneuvering around and how much energy is consumed and so forth, maybe she would start to be aware of these tradeoffs and would jump in on some of these other things.
And, again, that's getting back to this quantified self-metaphor. The idea is as you see your behavior, you're aghast that, you know, oh, my God, I consume 400 calories of chocolate chip cookie every day. I've gotta cut that out right away.
Or, you know, the average is 10,000 steps, you know, a day and I don't even get that in a week. Something's seriously out of whack.
I'm a strong believer that confronted with the data, people will be able to make the right decision or that clever people will produce apps and services that exploit those data to help you go on to make the right decision. And this is more kind of in the line of nudge and things like that. It's not always obvious to you what the right thing is, but given enough data, others might be able to design, you know, programs and applications that let you move forward.
So here are a few people that I want to recognize, my collaborators. And, actually, there should be two more there for Tom Defanti and Larry Smarr at San
Diego.
This is our generic Michigan interactive and social computing site. So not only is this work represented there but the work of all of my colleagues, some of which is strongly related, like the stuff that Resnick is doing with personal health and some of the work that [inaudible] is doing and so forth. And then these are the awards from the National Science Foundation, which we're quite grateful for, and the University of Michigan Office of the Provost has also been a sponsor of this work.
So I guess we're at the end of the hour, but if others have questions, I'm happy to entertain them.
Thank you.
[applause]
>> Tom Finholt: Yes?
>>: I have an observation.
>> Tom Finholt: Okay.
>>: A personal conflict [laughter] which leads into sort of your study, up coming study. So the observation is if you look back in history, we've moved from on when needed to always on, right? So the telephone being kind of an example of that where my mom still has a rotary phone. It still works, interestingly enough.
Why, I have no idea. But there's still some -- in various cities some. And it only uses electricity when you need it in very small amounts, right? And so -- versus my phone which is always on. And so I was thinking about the fact that we're going to introduce all these monitoring devices which themselves are always on, consuming electricity, right? Which is like, okay, we're just kind of -- and then I think, using my mom as a metaphor again, it's sort of -- she's got the rotary phone with, you know, the cord. So she's got to walk to the phone, which means that she gets a little bit of exercise along the way, right? Which goes back to the counting. She does it naturally without having to -- she has to walk to that phone in order to do that. If she's downstairs, she walks up the stairs to the phone versus me just pulling it out of my pocket. So she's always, you know, just by life, the way things are, she's naturally -- and very, very efficient. She's probably -- she's far more efficient than anybody else. Even if I were to put a solar panel in the house, she's going to be more efficient than me just from her habits and practices.
>> Tom Finholt: Just to flag that point, there was an article in the New York
Times this morning by a food writer who's been on a little campaign against fast food. And one of the things he noted is that over the last two, two and a half generations, that Americans have become -- it's not just fast food, it's all food production in general has become incredibly inefficient. And one aspect of that is we don't actually use very much of the beast when we eat the beast.
I mean, it used to be that every bit of the beast was processed and you had tongue and you had tripe and you had all that other stuff which, you know, my kids never even heard of. My wife says that her mom in western Michigan had a big jar of tongue on the counter all the time, and that was a staple for sandwiches.
So I think this plays into this general theme that there's -- it's not slop as much as it is just like abundance, yeah.
>>: And that's segues sort of to the issue of your story, and I'm interpreting it correctly, and that is that you're going to look for those early adopters to pick up these monitoring equipment. Naturally they're going to be -- like the Prius. When you see people with Priuses looking to -- oh, if I accelerate a little bit slower, what's going to -- there's sort of this biofeedback mechanism automatically.
>> Tom Finholt: Right.
>>: So, you know, you have an interesting dilemma here where you have the
Heisenberg -- the effect of measuring is going to have such a huge effect on their
behavior --
>> Tom Finholt: Right.
>>: -- so I think you're going to have a fair amount of difficulty resolving that.
>>: That's the goal, though, isn't it?
>>: I'm going to say optimize for the wrong thing. So I could win at this game by buying a new washing machine or new refrigerator, and that's true in the short run. It might not be true if you looked at it -- so I think you just need to worry about what you measure and what behavior that's encouraging and whether that's what you want for the long run.
>> Tom Finholt: Yeah. I mean --
>>: [inaudible] new practices and old practices.
>>: No, you're right. The Prius gauge didn't affect my behavior. But on the other hand, I suppose if you get in one more accident, you know, every month in
Seattle because people are looking at their energy gauge or saying, oh, if I just don't brake right now, look how much energy --
>>: That's what they do --
>> Tom Finholt: But that's actually a good point. And there are --
>>: Like a [inaudible] and everybody's always in the hospital with pneumonia.
>> Tom Finholt: Right. Or whatever, some other unintended or secondary -- I mean, technology is filled with these kind of secondary consequences, like use of antibiotics and so forth.
That to me is actually a somewhat heartening story because it says that there is still a place for design and human computer interaction even as these capabilities proliferate and become embedded in the other technologies.
And I think there has been some sign that [inaudible] has moved in that direction, principally as an influence of the work going on in [inaudible] group and in
[inaudible] group and so forth, but there's others around the country that have sort of taken up that banner. And I think that's a really interesting take to start to say -- what do they call it -- sustainable HCI, I think, or something like that.
>>: [inaudible]
>> Tom Finholt: Yes. [inaudible] and you would be another example. Carnegie
Mellon.
>>: Yeah, I think this is great. All the time I've had these questions like the one about computers, turning them off or on, or another one is which is better, to print
20 copies of something or to print one and photocopy it. And nowadays they're all the same machine, but they used to be -- and so maybe it doesn't matter at all or maybe --
>> Tom Finholt: So the one that we kind of spin over on is, is it better to know
100 people at sort of a fairly lightweight level and consume whatever little bits of energy are out there and that produces some social utility for me that is some function of that, or should I actually get in the plane and expend the jet A and go and have an intense week with you and, you know, that supercharges me to go back into my shop at home and I invent 20 new technologies that make the world better in a hundred different ways? And that's the kind of thing that we're trying to struggle with here.
And it's a little bit nerdy, I think, to try to quantify the nature of these bonds, but the point is that we don't actually right now. I mean, a lot of social network analysis is really a tie exists or it doesn't. There's not as much effort to sort of try to characterize the nature of it in terms of your well-being. My network analytics friends would stroke out to hear me say this, but, I mean, it's a lot of -- Mark
Newman when he does his analyses, he flat out says I don't actually care what the nodes and the edges are, you know. As a physicist, you know, is it molecular reactions, and grade school kids, is it high schoolers dating? I don't actually care. I'm interested in the algorithms.
But I actually care. I'm interested in what's going on there. And I think these kinds of tools are going to give us some insight into those questions.
>>: I think there's some -- I mean, another one that I've heard people arguing about when I was on the elevator was two people arguing about how much energy does using the elevator take? One person says, oh, it uses a lot of energy, and the other person says, no, it's all counterweight balanced and counterweighted and it hardly uses any at all. Well, if you hung a little sign in the elevator that said -- you know, this kind of metaphor, I think that would change some people's behavior, unless, of course, it takes almost none. But do I think -- so I do think -- I think getting the data out there will change behavior.
>> Tom Finholt: Well, the main thing is if I fore go a European trip, let's face is,
Delta's flying that plane over the ocean whether I go or not. And actually all of us, even whoever is watching remotely, if we all decided not to take those trips,
Delta is still flying that plane over the ocean.
But if there's some more profound shift, then they --
>>: Have one fewer flight.
>> Tom Finholt: Yeah. They'd build smaller planes or they actually -- they realize that we care about the -- which is not to say the airline industry isn't focused on efficiency because jet A is not getting --
>>: If they are concerned about packing the planes full, that's clearly probably one of the bigger gains for transportation --
>> Tom Finholt: Right, but at the expense of the travel public. And there's another one of those things where, you know --
>>: [inaudible]
>> Tom Finholt: -- does the packing -- you know, what incidence of viral contagion is amplified by the higher density. We can play these games --
>>: [inaudible]
>> Tom Finholt: All right. Thanks.
>>: I'll see you over the semester.
>> Tom Finholt: Definitely. See you later.
>>: -- which was kind of fun where they -- there was a group of woman that are very adamant about having child seats for infants, totally [inaudible] totally horrific and all that stuff, and, well -- so they're pushing hard. Now, the consequence of it is that you have the blankets for a car seat. But the consequence of that is a lot of families wouldn't want to do that expenditure which is drive, which is to drive, the chances of getting hurt or killed are astronomically higher than being in that plane and being in an air pocket and suddenly hit the ceiling, right? So here is this sort of passion dictating, you know, desired behavioral change to think that the consequences would be totally negative in the grander scheme of things. But the up front visualization of your baby hitting the ceiling is so stark and strong that --
>> Tom Finholt: Yeah, Freakonomics talks about that, that some things are so visceral, like the kidnapping of a child, that you're just -- you can't be rational about it. So I read -- there's this woman that writes this book about free-range children. She lets her -- basically probably how we were when we were kids.
You leave the house at seven and you return at nine p.m. She's saying that given the actual risk of abduction, a child would have to stand out on its front lawn for something like a million years before something bad would happen, and yet the horror of that is so visceral that it drives a host of related behaviors. Like do kids walk to school anymore?
>>: Yeah, that one -- yeah, that one -- that's completely true for me. That completely is -- I just can't get that image out even though I was one of the free-range kids, and yet it's just really hard for me. It's because of those news stories. That one is a very powerful one.
>> Tom Finholt: Which is to say that if you could implant some of these
[inaudible] into society, I am pretty optimistic that you could have dramatic consequences. So clearly this [inaudible] of the abduction has now altered the behavior of a generation of kids. And as an English teacher friend of mine in Ann
Arbor High School was -- did I tell you this story? He was teaching Tom Sawyer, and he asked his students how old did they think Tom Sawyer was, and they all
said he must be 17 or 18. Well, why do you think that? Well, because he's going everywhere on his own. He never has to ask permission. He's completely independent. And from there point of view that was impossible. So things could change.
>>: I've wondered how [inaudible] do you have to go before it's like it was
[inaudible].
>>: I grew up in New York and I was reasonably free-range too.
>>: That's true. When we were in Boston I was free-range.
>>: I mean, we could -- in fact more so than that, so as a 11- or 12-year-old we would take the train to the stadium to watch a game.
>>: Yeah, that's right.
>>: And we did it on our own because our parents -- we didn't have a car, so we would take, you know, the train. It was 25 cents and then --
>>: Yeah, you're right. That's what I did in Boston at the same age. My younger brother did it too. We just went anywhere we wanted on the subway.
>>: So I don't know. Rural or not, I think it was just different times
>> Tom Finholt: My dad was talking with my kids about the end of World War II, and he said on VJ day he got on the train in Chicago and went down just to see the crowds, and my oldest son started to do the math and said, you were 11 when you did that. You were gone all day? Weren't your parents concerned about you? No.
>>: My grandfather in New York City dropped out after the second grade to get a job as a messenger boy running stuff around downtown New York City, and so he was even younger and he was working. I mean, it was --
>>: A lot of the time, too, you were expected to be more mature faster, right? I mean, my folks eventually -- at age 14 school had ended and they went to work at 14, where now it's maybe you're 26 [laughter].