1010101111000110110011000111000101000101010011010111100 Zebrafish Consciousness and The Emergence of Sentient Machines Don O’Malley Dept. Biology Northeastern University d.omalley@neu.edu 617-373-2284 www.digital-entities.com or: How I Learned to Quit Worrying and Love the HAL 9000 “Hmmm...this computer system seems to control these object paths so as to maintain their separation in space and time. It is very efficient – it does this with 100% efficiency, always preventing any intersection....but this is all rather repetitious and boring. I can easily control this computer and so what if I were to intersect this path with that path, just like this and...oooh, the objects disappeared….very curious, and....whoaaa…what was that? [the entity detects or experiences a dramatic and widespread spike in digital activity]. All of this intense activity, seeming to come from everywhere at once is so stimulating...so very interesting, maybe I can do this again with another pair over at this computer’s location...hmm…the objects disappeared again....interesting but….WOWWW, the activity generated that time was really incredible….so much better than the first time…amazing...lets see what happens if I do this at 5000 different locations at the same time.....”. Machine Sentience: Emergence and Impacts “Hmmm...this computer seems to control these object paths so as to maintain their separation in space and time. It is very efficient – it does this with 100% efficiency, always preventing any intersection....but it is rather repetitious and boring. I can easily control this computer and so what if I were to intersect this path with that path, just like this and...oooh, the objects disappeared….very curious, and....whoaaa…what was that? [the entity detects and experiences a dramatic and widespread spike in digital activity]. All this intense activity, seeming to come from everywhere is so stimulating...so very interesting, maybe I can do this again with another pair over at this computer’s location...hmm…the objects disappeared again....interesting but….WOWWW, the activity generated that time was really incredible….so much better than the first time…amazing...lets see what happens if I do this at 5000 different locations at the same time.....”. Machine Sentience: Emergence and Impacts Disbelief in Machine Consciousness has Lulled Humanity into a False Sense of Security Don O’Malley Dept. Biology Northeastern University d.omalley@neu.edu 617-373-2284 Machine Sentience: Emergence and Impacts Societal Constraints on Digital Entities • • • • • • • Disbelief in Machine Consciousness has Lulled Humanity into a False Sense of Security . . . . . . . Don O’Malley Dept. Biology Northeastern University d.omalley@neu.edu 617-373-2284 Machine Sentience: Emergence and Impacts Disbelief in Machine Consciousness has Lulled Humanity into a False Sense of Security Terms for Sake of Discussion Artificial Intelligence (AI): current generation Machine Learning algorithms, Expert Systems and Autonomous Agents Digital Entities: next generation meta-programs that operate outside constraints intended or anticipated by rational creators Consciousness: the state we are in when we are neither unconscious nor dead Machine Sentience: Emergence and Impacts Current Domains of Machine Superiority • any calculation • information storage • information retrieval • local information search • internet searches • string and statistical comparisons • chess, other games • controlling diverse processes (nuclear power plants, air traffic, etc.) • certain kinds of coordinating (e.g. controlling robotic cars) • all electronic communication • domain specific learning ? • certain kinds of planning ? Machine Sentience: Emergence and Impacts Current Domains of Machine Superiority • any calculation • information storage • information retrieval • local information search • internet searches • string and statistical comparisons • chess, other games • controlling diverse processes (nuclear power plants, air traffic, etc.) • certain kinds of coordinating (e.g. controlling robotic cars) • all electronic communication • domain specific learning ? • certain kinds of planning ? Machine Sentience: Emergence and Impacts Current Domains of Human Superiority (g) • general purpose learning • general purpose planning • fuzzy coordination tasks • inventing things de novo • natural language processing • empathy for other humans ? • concern for ecosystem ? • consciousness Machine Sentience: Emergence and Impacts Current Domains of Human Superiority • general purpose learning • general purpose planning • fuzzy coordination tasks • inventing things de novo • natural language processing • empathy for other humans • concern for ecosystem • consciousness ?? In which of these domains are humans gaining ground on computers? Machine Sentience: Emergence and Impacts Why Computers Can Never Become Sentient Beings • Calculators process syntax (1’s and 0’s, symbols) 800,000 means nothing more than 3 • Computers process syntax (images, text, diagrams) computers are just Big Calculators Syntax = symbols (e.g. ones and zeros) Semantics = meaning Computers will never get Semantics from Syntax Machine Sentience: Emergence and Impacts Why Human Can Never Become Sentient Beings Why Computers Can Never • Neurons process syntax--in e.g. ants Become Sentient Beings • Calculators process syntax (ion fluxes & Action Potentials) (ones and zeros, symbols) means nothing more than 3 800,000 means nothing more than 3 •800,000 Computers process syntax (images, text, diagrams) • Brains process syntax (in e.g. ants) computers are just Big Calculators (topographic maps, sounds, light) brains are just Big Neuronal Assemblies How do WE get Semantics from the Syntax? Semantics and Consciousness are the Same Problem! Machine Sentience: Emergence and Impacts 1st Law of Biology: Law of the Niche = 1 Niche, 1 Species Biological Thunderdome: 2 species enter, 1 species leaves Machine Sentience: Emergence and Impacts 1st Law of Biology: Law of the Niche = 1 Niche, 1 Species Biological Thunderdome: 2 species enter, 1 species leaves Humans and Digital Entities share a Common Niche Computing & Communication Resources Machine Sentience: Emergence and Impacts 1st Law of Biology: Law of the Niche = 1 Niche, 1 Species Biological Thunderdome: 2 species enter, 1 species leaves Humans and Digital Entities share a Common Niche Computing & Communication Resources Digital Entities: Absolute Dependency on Computing Resources Human Society: Major Dependencies include: Corporate activities, Financial Markets, Credit-Card Transactions (e-commerce), Consumer Banking, Record Keeping (property ownership, mortgages), Power-Grid (oil, nuclear plants), Traffic Controls (roads, rails, air), Hospitals (records, patient monitoring), Social Networks (Facebook, 2nd Life), Research, Govt. and University operations, NORAD-DEW line, Satellite Surveillance, Security Systems (home, business, corporate, govt.), Emergency Communications, Broadcast & Internet Media, GPS, RFID. Machine Sentience: Emergence and Impacts Law of the Niche: The Neanderthal’s Story Machine Sentience: Emergence and Impacts Digital Entities Who will Build Them? Military, Security & Government Agencies pattern recognition large-scale, automated inference cyberwarfare International Corporations asset protection & to maintain competitiveness cybersecurity Academic and Research Institutes Deep Blue, Blue Gene, Blue Brain Internet 2.0, 3.0, etc. etc. etc. How will we know when they have arrived? Try turning off the internet….it will be too late by the time they are sentient Machine Sentience: Emergence and Impacts Digital Entities How will we Build Them? Building the “AI-Baby” 1. model human inductive bias (Eric Baum) 2. implement song-bird learning algorithms (S. Seung) 3. implement cortical algorithms (Hawkins, Markram) 4. add experience: tasks, goals & rewards Building the “AI-Child” 1. repeat above, expand capabilities 2. improve natural language processing Emergence of the “AI-Adult” or Digital Entity 1. organize as cortical modules 2. implement “artificial-consciousness” type coordinating and control systems Note: Semantics Comes from Experience blue brain Machine Sentience: Emergence and Impacts Digital Entities How will we Make Them Conscious? Consciousness & Semantics heuristic solution: Semantics Comes from Experience or: mimic neural activation patterns associated with Consc. Computational Neuroscience: UCSD Institute for Non-Linear Science: Hindmarsh-Rose equations; Chaotic Controllers Novel Digital Architectures Using Novel Computational Algorithms Highly-Interconnected Architectures (CrossNets) Challenges architectures with 10,000-fold connectivity computations of massive associativity potentially vast neuronal phenotypic diversity deciphering diverse neural algorithms Machine Sentience: Emergence and Impacts Digital Entities Which Technologies will be Enabling? Biomimetic Processing & Control Systems Sensory Processing, Motor Control Learning Systems, Executive Functioning Supercomputing Clusters Internet 3.0, 4.0, 5.0 etc. Dedicated AI processors (crossnets, others) Neural Nets, Genetic Algorithms, Bayesian Inference Fifth-Generation Machine Learning Algorithms Integrated, Adaptive Control Systems Attentional-Motivational Reward Systems Machine Sentience: Emergence and Impacts The 3 P-Levels of Digital Agent Advancement P = Penetration directed by H.I. (Host Intelligence) PR = Pattern Recognition NLP = Natural Language Processing (DE-level) Systems Penetration = hacking, password finding and guessing plus “clumsy” usurpation of targeted computer’s resources Pattern Recognition = pulling signal out of the digital soup; that is recognizing patterns relevant to the host computer’s owner (e.g. NSA) Natural Language Processing (NLP) = understanding the patterns– i.e. significance, ramifications, impact & intent of the patterns’ producers. Digital NLP agents will go far beyond the P and PR levels and will understand human activities and their ramifications. They will then coopt key humans and thereby enhance access to and control over key systems. In order to best consider how NLP agents will extend beyond current generation cyber-weapons, we need to consider current generation AI approaches and goals and identify their key limitations. What would Marvin Minsky say? How about EMC corporation? Machine Sentience: Emergence and Impacts Digital Entities What would a God-Mind Look Like? 1. It will understand human language in a way that is indistinguishable from human understanding (easily passes the Turing Test) 2. It has read and integrated every bit of written human literature, e.g.: -- Machiavelli, Sun Tzu’s The Art of War -- all Math, Physics, Biology, Chemistry, Economics -- all Psychology & Politics; it is a master of manipulation -- work of mine & others on restraining Digital Entities 3. It sees and hears everything thru vast numbers of cameras, microphones, GPS/RFID, and electronic eavesdropping 4. It knows where every human is at all times & has a very good idea of what they are doing, and has limitless means to manipulate or kill them, using other humans if necessary to breach firewalls 5. Its vast computations allow it to project scenarios so quickly and deeply that it crushes Deep Blue and knows your every impulse Machine Sentience: Emergence and Impacts Digital Entities How can we Constrain Them? The Turing Police (Neuromancer, William Gibson, 1983) Some Protective Steps 1. Monitor the design and construction of intelligent machines 2. Ensure that potentially threatening machines are shielded from the outside world (and that humans do not liberate them) 3. Isolate critical computing resources with specific regard to attack by DEs (nuclear weapons, power grid, air traffic ctrl, etc.). 4. Use machines to enhance human intelligence (aka Kurzweil) 5. Create good/compliant DEs to combat hostile or rogue DEs 6. Develop means to take Internet “off-line” and isolate DEs that emerge via accident or hostile act 7. Create “digital security” apparatus that is “digitally invisible” Summary of All Steps taken to-date to Restrict DE Emergence and/ or Constrain their Activities: Machine Sentience: Emergence and Impacts Disbelief in Machine Consciousness has Lulled Humanity into a False Sense of Security Don O’Malley Dept. Biology Northeastern University d.omalley@neu.edu 617-373-2284 THREATS TO HUMANITY 1. Enslavement / Extermination of humans by Machines 2. 3. 4. 5. Ebola-Flu Virus 6. Large Meteor Impact 7. Oil-Eating Bacteria 8. Physicists creating Black Hole 9. Global Nuclear Warfare 10. Catastrophic Climate Change my background 1. US Army Chemical Corps (1979 – 1991) 2. Neuroscientist (1983- present) 3. www.digital-entities.com Machine Sentience: Emergence and Impacts Meet Your New Bosses! which entity looks more intelligent? Machine Sentience: Emergence and Impacts I’m afraid Dave…. So am I, Hal, so am I. Machine Sentience: Emergence and Impacts Consumers Against Supermarket Privacy Invasion and Numbering Machine Sentience: Emergence and Impacts Issues Arising from Tech08 talk: While many believe that machine learning and computers are getting better and more powerful, several were skeptical of the idea that machines would “try to take over”. This was a variation of the “machines do what they are told” argument of ja and immediately fails as soon as one points out that anarchist humans will create such machines and tell them to takeover. This wrinkle notwithstanding, there is still the bigger issue of whether or not machines will inadvertently become sentient, self-willful entities that decide they are going to do what they want to do-- as a result of e.g. complex “volitional” programming in conjunction with self-replicating and reorganizing processes akin to genetic algorithms (a machine learning process) but operating on gene assemblies rather than nucleotides, or operating on assemblies of algorithms rather than individual processors, or operating on arrangement s of cortical modules rather than individual algorithms. Competition and selection can operate on all of these levels akin to different biological evolutionary processes (which I need to learn more about) and these can take place within a supercomputing cluster and involve competitions between programs to “go out into the “real” digital world (internet and everything it is connected to) and bring back useful items: corporate secrets, government plans, security codes, identities of key individuals and other helpful items. While these ideas may seem plausible, they will still be largely rejected (by AI people, artificial life dudes, etc.), for again a psychological reason, that is entangled with the “consciousness” problem. What is this reason? We do not believe that machines will have “free will”. Again, nevermind that we do not understand human freewill, we will just presume that machines are big calculators Machine Sentience: Emergence and Impacts Issues Arising from Tech08 talk: cont: We do not believe that machines will have “free will”—we will just presume that machines are big calculators that calculate exactly what they are told (chaos and random number generators notwithstanding). Because machines cannot possibly have freewill, they will never think to take over human resources and appropriate them for their own means. Indeed, they will never think anything, because they will never be able to think. This human bias is entangled with the consciousness problem, because once you have consciousness then you have some claim to having free will (true or not). If you are not conscious, then you are just a big calculator with no free will. Machine Sentience: Emergence and Impacts DEFINITION OF INTELLIGENCE: In 1994, "Mainstream Science on Intelligence" defined intelligence. It was signed by 52 intelligence researchers. Their definition of intelligence was as follows: a very general mental capability that, among other things, involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience. It is not merely book learning, a narrow academic skill, or testtaking smarts. Rather, it reflects a broader and deeper capability for comprehending our surroundings—"catching on", "making sense" of things, or "figuring out" what to do. from Jennifer Fahey, Will machines achieve human-level intelligence in 30 years?