Machine Learning - Bryant McGill

# Machine Learning **Entity class:** Concept or analytic term > **Machine-Evolution Nexus:** [[articles/Digital Darwinism and the Invisible World of Machine Evolution|Digital Darwinism and the Invisible World of Machine Evolution]] connects this node to the substrate-independent history of heredity, selection and computational populations. **Domain:** Artificial Intelligence / Statistics / Adaptive Systems **Doc Type:** Canonical Discipline Node **Maturity:** Foundational ## Definition **Machine learning** develops systems whose behavior or predictions improve through exposure to data, feedback or search rather than through exhaustive specification of every decision rule. ## Evolutionary Nexus [[articles/The Evolutionary Roots of Silicon Valley|The Evolutionary Roots of Silicon Valley]] places machine learning inside a longer history of adaptive inference: paleontology reconstructs systems from fragments, genomics reads descent from sequence, expert systems encode biological expertise and evolutionary computation searches by variation and selection. Not every machine-learning method is literally Darwinian. Gradient descent changes parameters along a calculated local direction; supervised learning minimizes error against labeled examples; reinforcement learning updates behavior under reward; evolutionary algorithms operate on populations with explicit variation and selection. Their commonality is differential retention under an evaluative structure, not one identical mechanism. [[Arthur Samuel|Arthur Samuel’s]] [[Samuel Checkers|IBM checkers programs]] provide an early, concrete waypoint. They used evaluation, selective search, accumulated play, and retained experience to improve inside a bounded environment. The work began before the [[IBM Thomas J. Watson Research Center]] opened, but belongs to the [[IBM Research]] lineage Yorktown later consolidated. ## Continuity Context Learning systems become part of human cognitive development when they mediate memory, attention and decision. This makes model updates a form of environmental change within [[wiki/Human-Machine Symbiosis|Human-Machine Symbiosis]], with duties of disclosure, contestability and preservation of user agency. ## Key Insight **Machine learning does not eliminate design; it relocates design into data, objectives, representations, feedback and the environment in which adaptation occurs.** ## Symbolic Language Engine Context [[projects/Ten Years Building a Symbolic Language Engine|Ten Years Building a Symbolic Language Engine]] establishes an explicit symbolic baseline for a modern hybrid system. [[wiki/Word Embeddings|Word embeddings]], [[wiki/Vector Retrieval|vector retrieval]], neural pronunciation models, and local [[wiki/Language Model|language models]] could enlarge the candidate space without erasing the older system's distinctions among phonological, semantic, taxonomic, corpus-attested, and formally constrained results. That division of labor is the core of [[wiki/Neuro-Symbolic AI|neuro-symbolic AI]]: learning supplies reach and adaptability; explicit representations supply controllable constraints, interpretable relations, and provenance. ## Simple Reminders, Quotations, and Thoughts > "I'm thinking about the difference between artificial intelligence and artificial life." > **— Esther Dyson**, *2015, Edge annual question “What Do You Think About Machines That Think?”* [[reminders/Machine Succession/Artificial Intelligence and Artificial Life Are Different by Esther Dyson|Artificial Intelligence and Artificial Life Are Different by Esther Dyson]] > "The speed of the fastest birds did not turn out to be a limit to airplanes, and artificial minds will be faster, more accurate, more alert, more aware and comprehensive than their human counterparts." > **— Joscha Bach**, *2015, Edge annual question “What Do You Think About Machines That Think?”* [[reminders/Machine Succession/Artificial Minds Will Surpass Biological Limits by Joscha Bach|Artificial Minds Will Surpass Biological Limits by Joscha Bach]] > "A common theme in recent writings about machine intelligence is that the best new learning machines will constitute rather alien forms of intelligence." > **— Andy Clark**, *2015, Edge annual question “What Do You Think About Machines That Think?”* [[reminders/Information/Learning Machines May Develop Alien Intelligence by Andy Clark|Learning Machines May Develop Alien Intelligence by Andy Clark]] > "Looking ahead, I could live with a partnership with machine learning in order to make complex modern life more resource-efficient in a way that human brains cannot." > **— Laurence C. Smith**, *2015, Edge annual question “What Do You Think About Machines That Think?”* [[reminders/Machine Succession/Machine Learning Could Make Modern Life More Efficient by Laurence C. Smith|Machine Learning Could Make Modern Life More Efficient by Laurence C. Smith]] > "Machines can faithfully imitate the results of some human thought processes whose outcomes are fixed (remembering people's favorite movies, recognizing familiar objects) or dynamic (jet piloting, grand master chess play)." > **— Scott Atran**, *2015, Edge annual question “What Do You Think About Machines That Think?”* [[reminders/Memory/Machines Can Imitate Fixed and Dynamic Human Thought by Scott Atran|Machines Can Imitate Fixed and Dynamic Human Thought by Scott Atran]] > "There is also exceptionally high potential for applications of transfer learning. So much of the practical value of machine learning, for example in search and information retrieval, has traditionally focused on systems that learn from the massive datasets and people available on the World-Wide Web. But what can web-trained systems learn about smaller communities, organizations, or even individuals? Can we foresee a future where intelligent machines are able to learn useful tasks that are highly specialized to a specific individual or small organization? Transfer learning opens the possibility that all the intelligence of the web can form the foundation of machine-learned systems, from which more individualized intelligence is learned, through transfer learning. Achieving this would amount to another step towards the democratization of machine intelligence." > **— Peter Lee**, *2017, Edge Annual Question, “What Scientific Term or Concept Ought to Be More Widely Known?”* [[reminders/AI Control/Transfer Learning Could Democratize Personalized Machine Intelligence by Peter Lee|Transfer Learning Could Democratize Personalized Machine Intelligence by Peter Lee]] > “Imagine an artificial intelligence, with human-like insight, contemplating her own blueprint. What would she make of it? I think it's overwhelmingly likely that among her first thoughts would be how to begin making improvements. This processor could be faster, that memory more capacious—and, above all, the reward system more rewarding!” > **— Frank Wilczek**, *2014, Edge Annual Question, “What Scientific Idea Is Ready For Retirement?”* [[reminders/AI Control/A Self-aware AI Would Immediately Redesign Itself by Frank Wilczek|A Self-aware AI Would Immediately Redesign Itself by Frank Wilczek]] > “The sectors include mobility, sensors, cloud computing, and data analysis, whether by machine learning or artificial intelligence. Sensors don’t just give us new information about nature and society, they inform the configuration of cloud systems, and the behavior of the analysis algorithms is likewise affected by the success with which it alters the other two.” > **— Quentin Hardy**, *2016, Edge Annual Question, “What Do You Consider The Most Interesting Recent [Scientific] News? What Makes It Important?”* [[reminders/AI Control/AI Sensors and Cloud Systems Are Converging into One Feedback Loop by Quentin Hardy|AI Sensors and Cloud Systems Are Converging into One Feedback Loop by Quentin Hardy]] > “The first class of stories is about the science, where many researchers are now vocally pointing out that there is a lot more science to be done in order to come up with learning algorithms that mimic the broad capabilities of humans and animals. Deep learning by itself will not solve many of the learning problems that are necessary for general Artificial Intelligence, for instance where spatial or deductive reasoning is involved. Further, all the breakthrough results we have seen in AI have been years in the making, and there is no scientific reason to expect there to be a sudden and sustained series of them, despite the enthusiasm from young researchers who were not around of the last three waves of such predictions in the 1950's, 1960's, and 1980's.” > **— Rodney A. Brooks**, *2016, Edge Annual Question, “What Do You Consider The Most Interesting Recent [Scientific] News? What Makes It Important?”* [[reminders/AI Control/AI Still Needs New Scientific Breakthroughs by Rodney A. Brooks|AI Still Needs New Scientific Breakthroughs by Rodney A. Brooks]] > “As for our place in nature, so too for our relationship with technology. Recent progress in artificial intelligence and bionics, in particular, have led to a great deal of soul-searching about who—or what—is in charge, and even what it means to be human. The industrial revolution saw machines replace human physical labor, but now that they are replacing mental labor too, what will be left for people to do? Even those who don't fear for their jobs might be angry when they discover that their new boss is an algorithm.” > **— Timo Hannay**, *2016, Edge Annual Question, “What Do You Consider The Most Interesting Recent [Scientific] News? What Makes It Important?”* [[reminders/Machine Succession/AI and Bionics Are Redefining Who Is in Charge by Timo Hannay|AI and Bionics Are Redefining Who Is in Charge by Timo Hannay]] > “I believe that we will see within our lifetime the convergence of developments in artificial intelligence, knowledge representation, statistical grammar theories, and an emerging field — computational anthropology (informatic-based analysis and modeling of cultural values) — that will facilitate powerful new forms of machine translation to match the dreams of early pioneers of computation.” > **— Daniel L. Everett**, *2009, Edge Annual Question, “What Will Change Everything?”* [[reminders/AI Control/AI and Knowledge Representation Are Converging by Daniel L. Everett|AI and Knowledge Representation Are Converging by Daniel L. Everett]] > “For a long time my optimism centered on computing in general, and what kinds of benefits it might bring us. Events have shown I entertained far too modest an optimism—I'm embarrassed to say that the impact of the Internet, in particular the World Wide Web, eluded me completely at first. A few years ago, I returned to artificial intelligence, which I'd written about early on, and then gone away from. Press narratives were uncritical about the field's death throes, and I expected to write an elegy. Instead, I found a revelation. Artificial intelligence is not only robustly healthy, building on its very significant gains since I first wrote about it, but the field's present ambitions burst with, well, vitality.” > **— Pamela McCorduck**, *2007, Edge Annual Question, “What Are You Optimistic About?”* [[reminders/AI Control/Artificial Intelligence Returned with a New Set of Questions by Pamela McCorduck|Artificial Intelligence Returned with a New Set of Questions by Pamela McCorduck]] > “Soon we will be building not only Artificial Intelligence. We will be building Artificial Will. Systems with an ability to convert internal decisions and values into external change. They will be able to decide that they want to change the world. A plan inside becomes an action on the outside. So they will have to know what is inside and outside.” > **— Tor Nørretranders**, *2009, Edge Annual Question, “What Will Change Everything?”* [[reminders/AI Control/Artificial Intelligence Will Become Artificial Will by Tor Nørretranders|Artificial Intelligence Will Become Artificial Will by Tor Nørretranders]] > “Most importantly, the prioritized cultivation of high-quality training datasets might allow an order-of-magnitude speedup in AI breakthroughs over purely algorithmic advances. For example, we might already possess the algorithms and hardware that will enable machines in a few years to author human-level long-form creative compositions, complete standardized human examinations, or even pass the Turing Test, if only we trained them with the right writing, examination, and conversational datasets. Additionally, the nascent problem of ensuring AI friendliness might be addressed by focusing on dataset rather than algorithmic friendliness—a potentially simpler approach.” > **— Alexander Wissner-Gross**, *2016, Edge Annual Question, “What Do You Consider The Most Interesting Recent [Scientific] News? What Makes It Important?”* [[reminders/AI Control/Better Datasets May Accelerate AI by an Order of Magnitude by Alexander Wissner-Gross|Better Datasets May Accelerate AI by an Order of Magnitude by Alexander Wissner-Gross]] > “Next node foment can also be seen in areas of current conflict in scientific theories, where two elegant high-order paradigms with explanative power are themselves in competition, uncomfortable coexistence, or broken symmetry fomenting towards a larger explanatory paradigm. Some examples include a grand unified theory to unify the general theory of relativity with electromagnetism, mathematical theories that include both power laws and randomness, and a behavioral theory of beyond-human level intelligence that includes both computronium and aesthetics (e.g.; does AI do art, solely compute, or is there no distinction at that level of cosmic navel-gazing?).” > **— Melanie Swan**, *2012, Edge Annual Question, “What Is Your Favorite Deep, Elegant, Or Beautiful Explanation?”* [[reminders/Machine Succession/Competing Paradigms Can Foment a Larger Theory by Melanie Swan|Competing Paradigms Can Foment a Larger Theory by Melanie Swan]] > “To me, having my worldview entirely altered is among the most fun parts of science. One mind-altering event occurred during graduate school. I was studying the field of Artificial Intelligence with a focus on Natural Language Processing. At that time there were intense arguments amongst computer scientists, psychologists, and philosophers about how to represent concepts and knowledge in computers, and if those representations reflected in any realistic way how people represented knowledge. Most researchers thought that language and concepts should be represented in a diffuse manner, distributed across myriad brain cells in a complex network. But some researchers talked about the existence of a "grandmother cell," meaning that one neuron in the brain (or perhaps a concentrated group of neurons) was entirely responsible for representing the concept of, say, your grandmother. I thought this latter view was hogwash.” > **— Marti Hearst**, *2008, Edge Annual Question, “What Have You Changed Your Mind About? Why?”* [[reminders/AI Control/Computational Analysis of Language Requires Understanding Language by Marti Hearst|Computational Analysis of Language Requires Understanding Language by Marti Hearst]] > “Over the past few years, a raft of classic challenges in artificial intelligence which had stood unsolved for decades were conquered, almost without warning, through an approach long disparaged by AI purists for its "statistical" flavor: it's essentially about learning probability distributions from large volumes of data, rather than examining humans' problem-solving techniques and attempting to encode them in executable form. The formidable tasks it has solved range from object classification and speech recognition, to generating descriptive captions for photos and synthesizing images in the style of famous artists—even guiding robots to perform tasks for which they were never programmed!” > **— David Dalrymple**, *2016, Edge Annual Question, “What Do You Consider The Most Interesting Recent [Scientific] News? What Makes It Important?”* [[reminders/AI Control/Differentiable Programming Is Rewriting Artificial Intelligence by David Dalrymple|Differentiable Programming Is Rewriting Artificial Intelligence by David Dalrymple]] > “A growing number of physicists understand that the universe is not mathematical, but computational, and physics is in the business of finding an algorithm that can reproduce our observations. The switch from uncomputable, mathematical notions (such as continuous space) makes progress possible. Climate science, molecular genetics, and AI are computational sciences. Sociology, psychology, and neuroscience are not: they still seem to be confused by the apparent dichotomy between mechanism (rigid, moving parts) and the objects of their study. They are looking for social, behavioral, chemical, neural regularities, where they should be looking for computational ones.” > **— Joscha Bach**, *2016, Edge Annual Question, “What Do You Consider The Most Interesting Recent [Scientific] News? What Makes It Important?”* [[reminders/AI Control/Everything Is Computation by Joscha Bach|Everything Is Computation by Joscha Bach]] > “Lessons from evolutionary psychology indicate that developing specialized intelligences — artificial idiot savants — and networking them would achieve a mosaic AI, just as evolution gradually built natural intelligences. The essential activity is discovering sets of principles that solve a particular family of problem. Indeed, successful scientific theories are examples of specialized intelligences, whether implemented culturally among communities of researchers or implemented computationally in computer models. Similarly, adding duplicates of the specialized programs we discover in the human mind to the emerging AI network would constitute a tremendous leap toward AI. Essentially, for this aggregating intelligence to communicate with humans — for it to understand what we mean by a question or want by a request, it will have to become equipped with accurate models of the native intelligences that inhabit human minds.” > **— John Tooby**, *2009, Edge Annual Question, “What Will Change Everything?”* [[reminders/Evolution/Evolutionary Psychology Points Toward Mosaic AI by John Tooby|Evolutionary Psychology Points Toward Mosaic AI by John Tooby]] > “Fitness landscapes (sometimes called "adaptive landscapes") keep turning up when people try to figure out how evolution or innovation works in a complex world. An important critique by Marvin Minsky and Seymour Papert of early optimism about artificial intelligence warned that seemingly intelligent agents would dumbly "hill climb" to local peaks of illusory optimality and get stuck there. Complexity theorist Stuart Kauffman used fitness landscapes to visualize his ideas about the "adjacent possible" in 1993 and 2000, and that led in turn to Steven Johnson's celebration of how the "adjacent possible" works for innovation in Where Good Ideas Come From.” > **— Stewart Brand**, *2012, Edge Annual Question, “What Is Your Favorite Deep, Elegant, Or Beautiful Explanation?”* [[reminders/Evolution/Fitness Landscapes Explain Evolution and Innovation by Stewart Brand|Fitness Landscapes Explain Evolution and Innovation by Stewart Brand]] > “When reporters interviewed me in the 70's and 80's about the possibilities for Artificial Intelligence I would always say that we would have machines that are as smart as we are within my lifetime. It seemed a safe answer since no one could ever tell me I was wrong. But I no longer believe that will happen. One reason is that I am a lot older and we are barely closer to creating smart machines.” > **— Roger Schank**, *2008, Edge Annual Question, “What Have You Changed Your Mind About? Why?”* [[reminders/AI Control/Human-level AI Remains Farther Away Than Predicted by Roger Schank|Human-level AI Remains Farther Away Than Predicted by Roger Schank]] > “The list of existential threats to humanity includes climate change, nuclear war, pandemics, asteroid collisions and perhaps AI. And all of these can be avoided. Some can be addressed here on Earth. Others require activity in space, but with the ultimate aim of protecting the planet.” > **— Yuri Milner**, *2017, Edge Annual Question, “What Scientific Term Or Concept Ought To Be More Widely Known?”* [[reminders/Existential Risk/Intelligence Determines How Long Civilizations Survive by Yuri Milner|Intelligence Determines How Long Civilizations Survive by Yuri Milner]] > “No, I don't literally mean that we should stop believing in, or collecting, Big Data. But we should stop pretending that Big Data is magic. There are few fields that wouldn't benefit from large, carefully collected data sets. But lots of people, even scientists, put more stock in Big Data than they really should. Sometimes it seems like half the talk about understanding science these days, from physics to neuroscience, is about Big Data, and associated tools like "dimensionality reduction", "neural networks", "machine learning algorithms" and "information visualization".” > **— Gary Marcus**, *2014, Edge Annual Question, “What Scientific Idea Is Ready For Retirement?”* [[reminders/AI Control/Machine Learning Turns Data into New Forms of Understanding by Gary Marcus|Machine Learning Turns Data into New Forms of Understanding by Gary Marcus]] > “Textbooks in neuroscience, including one that I coauthored, say that memories are stored at synapses between neurons in the brain, of which there are many. In neural network models of memory, information can be stored by selectively altering the strengths of the synapses, and "spike-time dependent plasticity" at synapses in the cerebral cortex has been found with these properties. This is a hot area of research, but all we need to know here is that patterns of neural activity can indeed modify a lot of molecular machinery inside a neuron.” > **— Terrence J. Sejnowski**, *2005, Edge Annual Question, “What Do You Believe Is True Even Though You Cannot Prove It?”* [[reminders/AI Control/Memory Is Stored by Changing Synaptic Strength by Terrence J. Sejnowski|Memory Is Stored by Changing Synaptic Strength by Terrence J. Sejnowski]] > “Blind gathering of Big Data in biology continues apace, however, emphasizing transformational technologies such as machine learning—artificial neural networks, for instance—to find meaningful patterns in all the data. But no matter their "depth" and sophistication, neural nets merely fit curves to the available data. They may be capable of interpolation, but extrapolation beyond their training domain can be fraught.” > **— Roger Highfield**, *2017, Edge Annual Question, “What Scientific Term Or Concept Ought To Be More Widely Known?”* [[reminders/AI Control/Neural Networks Fit Curves to Data by Roger Highfield|Neural Networks Fit Curves to Data by Roger Highfield]] > “David Deutsch, a physicist at Oxford said: "No brain on Earth is yet close to knowing what brains do. The enterprise of achieving it artificially — the field of 'artificial intelligence' has made no progress whatever during the entire six decades of its existence." He adds that he thinks machines that think like people will happen some day.” > **— Roger Schank**, *2014, Edge Annual Question, “What Scientific Idea Is Ready For Retirement?”* [[reminders/AI Control/No Brain Yet Knows What Brains Do by Roger Schank|No Brain Yet Knows What Brains Do by Roger Schank]] > “Recording more and more images and corresponding brain patterns boosts the vocabulary in the individual’s visual dictionary of thought. Accuracy greatly increases with the quantity and quality of data and of the decoding algorithms. Jepsen has persuaded me that this is realisable within a decade, within the cost points of consumer electronics, and in a form that appeals to non-techies. Laborious techniques and huge, power-hungry, multi-million-dollar systems based on magnetic fields will be succeeded by optical techniques where the advantages of consumer electronics can really assert themselves; the power of AI algorithms will do the rest. This science-fiction future is not only realisable, but because of enormous potential benefits, will inevitably be realised.” > **— Peter Gabriel**, *2016, Edge Annual Question, “What Do You Consider The Most Interesting Recent [Scientific] News? What Makes It Important?”* [[reminders/AI Control/Open Water–the Internet of Visible Thought by Peter Gabriel|Open Water–the Internet of Visible Thought by Peter Gabriel]] > “John McCarthy, the late co-founder of the field of artificial intelligence, wrote, "He who refuses to do arithmetic is doomed to talk nonsense." It seemed incongruous that a professor who worked with esoteric high level math would be touting simple arithmetic, but he was right; in fact in many cases all we need to avoid nonsense is the simplest form of arithmetic: counting.” > **— Peter Norvig**, *2017, Edge Annual Question, “What Scientific Term Or Concept Ought To Be More Widely Known?”* [[reminders/AI Control/Refusing to Count Is an Invitation to Nonsense by Peter Norvig|Refusing to Count Is an Invitation to Nonsense by Peter Norvig]] > “Artificial intelligence researchers will, numerous experts attest, probably build systems that are "recursively self-improving"—that understand their own workings well enough to design improvements to themselves, thereby bootstrapping to a state of ever more unimaginable intellectual performance.” > **— Aubrey de Grey**, *2009, Edge Annual Question, “What Will Change Everything?”* [[reminders/AI Control/Self-improving AI Could Bootstrap Beyond Imaginable Intelligence by Aubrey de Grey|Self-improving AI Could Bootstrap Beyond Imaginable Intelligence by Aubrey de Grey]] > “"Deep Learning" algorithms are now showing us how to use artificial neural networks in ways that come closer than ever before to delivering learning on a grand scale. But we probably need "deep culture", as well as deep learning, if we are ever to press genuine hyper-intelligence from the large databases that drive our best probabilistic learning machines.” > **— Andy Clark**, *2013, Edge Annual Question, “What *should* We Be Worried About?”* [[reminders/AI Control/Super-a.i.s Won't Rule the World Unless They Get Culture First by Andy Clark|Super-a.i.s Won't Rule the World Unless They Get Culture First by Andy Clark]] > “My stomach hurts. I tell this to my wife and she suggests a medicine in the cabinet that she remembers I have used before and reminds me that it helped. Now, suppose that this was not my wife but a computer? Is it an ad? Does it matter? Can we do this. Yes. AI technology could easily employ models of people and there needs. (But, today, we are busy with key words.)” > **— Roger Schank**, *2016, Edge Annual Question, “What Do You Consider The Most Interesting Recent [Scientific] News? What Makes It Important?”* [[reminders/AI Control/Targeted Advertising Foreshadows Personalized AI by Roger Schank|Targeted Advertising Foreshadows Personalized AI by Roger Schank]] > “Neuroscientists have discovered that dopamine neurons, found in the brains of all vertebrates, are central to reward learning. The transient responses of dopamine neurons signal to the brain predictions for future reward, which are used to guide behavior and regulate synaptic plasticity. The dopamine responses have the same properties as the temporal difference learning algorithm used in TD-Gammon. Reinforcement learning was dismissed years ago as too weak a learner to handle the complexity of cognition. This belief needs to be re-evaluated in the light of the successes of TD-Gammon and learning algorithms in other areas of AI.” > **— Terrence J. Sejnowski**, *2007, Edge Annual Question, “What Are You Optimistic About?”* [[reminders/AI Control/The Brain Learns Like a Temporal Difference Algorithm by Terrence J. Sejnowski|The Brain Learns Like a Temporal Difference Algorithm by Terrence J. Sejnowski]] > “I believe (I know—but can't prove!) that scientists will soon understand the physiological basis of the "cognitive spectrum," from the bright violet of tightly-focused analytic thought all the way down to the long, slow red of low-focus sleep thought—also known as "dreaming." Once they understand the spectrum, they'll know how to treat insomnia, will understand analogy-discovery (and therefore creativity), and the role of emotion in thought—and will understand that thought takes place not only when you solve a math problem but when you look out the window and let your mind wander. Computer scientists will finally understand the missing mystery ingredient that made all their efforts to simulate human thought such naive, static failures, and turned this once-thriving research field into a ghost town. (Their failures were "static" insofar as people think in different ways at different times—your energetic, wide-awake mind works very differently from your tired, soon-to-be-sleeping mind; but artificial intelligence programs always "thought" in the same way all the time.)” > **— David Gelernter**, *2005, Edge Annual Question, “What Do You Believe Is True Even Though You Cannot Prove It?”* [[reminders/AI Control/The Cognitive Spectrum May Explain Why AI Cannot Simulate Human Thought by David Gelernter|The Cognitive Spectrum May Explain Why AI Cannot Simulate Human Thought by David Gelernter]] > “Anything simple enough to be understandable will not be complicated enough to behave intelligently, while anything complicated enough to behave intelligently will not be simple enough to understand.” > **— George Dyson**, *2004, Edge Annual Question, “WHAT'S YOUR LAW?”* [[reminders/AI Control/Intelligent Systems May Be Too Complex to Understand by George Dyson|Intelligent Systems May Be Too Complex to Understand by George Dyson]] > “Imperial China developed sophisticated technologies while neglecting science, and it is all too easy to imagine a society that embraces technology but represses science, until only technology remains. Or, one particular species of technology might achieve such dominance that it halts the advance of science in order to preserve itself.” > **— George Dyson**, *2014, Edge Annual Question: “WHAT SCIENTIFIC IDEA IS READY FOR RETIREMENT?”* [[reminders/AI Control/A Technology May Halt Science to Preserve Itself by George Dyson|A Technology May Halt Science to Preserve Itself by George Dyson]] > “Smart technology will never be really smart until it can decenter and anticipate.” > **— Gary Klein**, *2017, Edge Annual Question: “WHAT SCIENTIFIC TERM OR CONCEPT OUGHT TO BE MORE WIDELY KNOWN?”* [[reminders/AI Control/Smart Technology Is Not Smart Until It Can Decenter by Gary Klein|Smart Technology Is Not Smart Until It Can Decenter by Gary Klein]] > “Why are the errors that our best machine-learning algorithms make so different from the errors we humans make?” > **— Jason Wilkes**, *2018 Edge Annual Question, question* [[reminders/AI Control/Why Do Machine-Learning Errors Differ From Human Errors by Jason Wilkes|Why Do Machine-Learning Errors Differ From Human Errors by Jason Wilkes]] > “Learning, like many other attributes we thought only humans owned, turns out to be something we can program machines to do. Learning can be automated. While simple second-order learning (learning how to learn) was once rare and precious, it will now become routine and common. Just like tireless powerful motors, and speedy communications a century ago, learning will quickly become the norm in our built world. All kinds of simple things will learn. Automated synthetic learning won't make your oven as smart as you, but it will make better bread.” > **— Kevin Kelly**, *2016 Edge Annual Question, response passage* [[reminders/Machine Succession/Learning Is Becoming an Automated Feature of Things by Kevin Kelly|Learning Is Becoming an Automated Feature of Things by Kevin Kelly]] ## Sources / Provenance - Tom M. Mitchell, _Machine Learning_, 1997. - Christopher M. Bishop, _Pattern Recognition and Machine Learning_, 2006. - Historical synthesis: [[articles/The Evolutionary Roots of Silicon Valley|The Evolutionary Roots of Silicon Valley]]. ## See Also [[wiki/Supervised Learning|Supervised Learning]], [[wiki/Reinforcement Learning|Reinforcement Learning]], [[wiki/Gradient Descent|Gradient Descent]], [[wiki/Evolutionary Algorithms|Evolutionary Algorithms]], [[wiki/Objective Function|Objective Function]], [[wiki/Word Embeddings|Word Embeddings]], [[wiki/Vector Retrieval|Vector Retrieval]], [[wiki/Neuro-Symbolic AI|Neuro-Symbolic AI]], [[Arthur Samuel]], [[Samuel Checkers]], [[Machine Intelligence Continuum]]  ## Austin research connections [[wiki/Scientific Machine Learning|Scientific Machine Learning]] connects learned representations with physical models and uncertainty. The Austin cluster follows those methods across disease forecasting, nuclear-effects modeling, biometric security, and [[wiki/Digital Twin|digital twins]]. [[wiki/Multi-Use Inference and Control|Multi-Use Inference and Control]] distinguishes a documented research transfer from an architectural comparison and tests each use against its own data and objectives. **Research map:** [[wiki/Austin Executable Loop|Austin Executable Loop]] · [[research/The Austin Executable Loop|Master document]]  ## Relationships  This entry is routed through [[collections/War With Empire|War With Empire]] and [[collections/Neurotech|Neurotech]]. Its source context is developed in [[articles/2026 Annual Report on Brain-Computer Interfaces|2026 Annual Report on Brain-Computer Interfaces]], [[articles/Technologies for Consciousness Mapping and Transfer|Technologies for Consciousness Mapping and Transfer]], [[articles/The Next Interface Layer|The Next Interface Layer]], and [[articles/The Magic Kingdom and the Managed State|The Magic Kingdom and the Managed State]]. Status-qualified source edges are preserved in the terminal Research Edges section. ### Technology and research relationships - [[wiki/Disney Research|Disney Research]] is identified with the research domain **machine learning**.  - **Edge source route:** [[collections/Edge|Edge]] connects this topic to exact Annual Question passages promoted into the Simple Reminders archive. ## Research Edges  #### Interface and soft-sovereignty graph patch — 2026-09-10 **Resolved aliases:** `machine learning` **Collections:** [[collections/War With Empire|War With Empire]] · [[collections/Neurotech|Neurotech]] **Relationship source:** [[research/Interface and Soft Sovereignty Ecosystem Relationship Graph - 2026-09-10|Interface and Soft Sovereignty Ecosystem Relationship Graph — 2026-09-10]] **Related source articles:** [[articles/The Next Interface Layer|The Next Interface Layer]] · [[articles/The Magic Kingdom and the Managed State|The Magic Kingdom and the Managed State]] #### Incoming typed edges - **Edge 15 — [[wiki/Disney Research|Disney Research]] `research_domain` → this entry** — **CORPUS_V**; evidence `CORPUS_MAGIC`.