Ray Kurzweil Heats Up Artificial Intelligence Discussions

Stephen DeAngelis

November 23, 2012

Ray Kurzweil is a very bright and entertaining individual. He is also, at times, controversial. His new book, How to Create a Mind, has created a bit of stir in the artificial intelligence world. His genius isn’t being questioned; but, some of his conclusions are. Gary Marcus writes, “Ray Kurzweil is, by all accounts, a genius. He holds nineteen honorary doctorates, has founded a half-dozen successful companies, and was a major contributor to the field of artificial intelligence. … Time magazine recently featured Kurzweil on its cover, and Fortune described him as ‘a legendary inventor with a history of mind-blowing ideas.’ And now he has a new book, with a subtitle that suggests he has found another such idea: ‘How to Create a Mind: The Secret of Human Thought Revealed.'” [“Ray Kurzweil’s Dubious New Theory of Mind,” The New Yorker, 15 November 2012] It didn’t take long for people to react to Kurzweil’s latest ideas. Marcus, a professor of psychology at NYU, obviously has a few concerns about what Kurzweil writes. He continues:

“In the preface to the book Kurzweil argues, with good reason, that ‘reverse-engineering the human brain may be regarded as the most important project in the universe.’ He then presents a theory he calls ‘the pattern recognition theory of mind (PRTM)’ which he claims ‘describes the basic algorithm of the neocortex (the region of the brain responsible for perception, memory, and critical thinking).’ Kurzweil suggests that his conclusions are ‘inescapable’ and that the principles he espouses can be used ‘to vastly extend the power of our own intelligence.’ That would be big news. But does the book deliver? Kurzweil’s critics have not always been kind; the biologist PZ Myers once wrote, ‘Ray Kurzweil is a genius. One of the greatest hucksters of the age.’ Doug Hofstadter, the Pulitzer Prize winning author of ‘Gödel, Escher, Bach’ has been even harsher, saying once in an interview that ‘if you read Ray Kurzweil’s books … what I find is that it’s a very bizarre mixture of ideas that are solid and good with ideas that are crazy. It’s as if you took a lot of very good food and some dog excrement and blended it all up so that you can’t possibly figure out what’s good or bad.'”

From the headline of Marcus’ article, it’s clear that he believes there are more feces than findings in Kurzweil’s latest book. Marcus’s biggest objection is that Kurzweil writes more as a new-age sage than as a neuroscientist. Marcus reports, for example, that Kurzweil begins his discussion by offering “vague gestures toward an unusual kind of neuron called spindle cells, … but offers no references and very little direct evidence.” Marcus also takes umbrage with Kurzweil’s belief that the mind is machine-like. He writes:

“Kurzweil returns to the business of explicating and defending his main thesis—according to which the part of the brain that is most associated with reasoning and conscious thought, the neocortex, is seen as a hierarchical set of pattern-recognition devices, in which complex entities are recognized as a statistical function of their constituent parts. Kurzweil illustrates this thesis in the context of a system for reading words. At the lowest level, a set of pattern recognizers search for properties like horizontal lines, diagonal lines, curves, and so forth; at the next level up, a set of pattern recognizers hunt for letters (A, B, C, and so forth) that are built out of conjunctions of lines and curves; and at still a higher level, individual pattern recognizers look for particular words (like APPLE, PEAR, and  so on that are built out of conjunctions of letters). The acronym P.R.T.M., for Pattern Recognition Theory of Mind, is new, but to scientists in the field, the basic idea is significantly less new than Kurzweil’s subtitle (‘The Secret of Human Thought Revealed’) lets on. Anyone who knows the history of A.I. will recognize that the basic theory (and even the diagrams that are used to illustrate it) is very much in the spirit of a textbook model of vision that was introduced in 1980, known as neocognition.”

Marcus believes that the inescapability of Kurzweil’s concepts is undermined by the fact that Kurzweil didn’t bother “to build a computer model that instantiated his theory, and then compare the predictions of the model with real human behavior.” Ronald Bailey, the science correspondent for Reason Magazine, isn’t as offended as Marcus with Kurzweil’s arguments. He writes a rather favorable review of the book. [“Head in the Cloud,” Wall Street Journal, 16 November 2012] Bailey lays out Kurzweil’s argument for pattern recognition theory of mind, as did Marcus, but accepts it much more readily than Marcus did. He writes:

“The insight that brains are built of pattern-recognition modules leads Mr. Kurzweil to argue that it will be possible to design artificial intelligence in the much same way. ‘The next step, of course, will be to expand the neocortex itself with its nonbiological equivalent,’ he writes. These synthetic neocortexes, he says, will consist of vast parallel sets of pattern-recognition modules. Already by using online resources, Mr. Kurzweil notes, people are migrating their thinking and memories to the computational ‘cloud.’ And computation is getting ever cheaper and more pervasive; Mr. Kurzweil calculates that, later in this century, ‘a thousand dollars worth of computation will be trillions of times more powerful than the human brain.’ He foresees our augmenting our biological neocortexes by hooking them up wirelessly to cloud-based synthetic ones. These synthetic add-ons might be composed of trillions of pattern-recognition modules—a sort of transcendent iPhone. By around 2040, Mr. Kurzweil says, we will be able to replicate and upload the entire information content of our brains into the cloud.”

Marcus counters, “Does the P.R.T.M. predict anything about human behavior that no other theory has predicted before? Does it give novel insight into any long-standing puzzles in human nature? Kurzweil never tries to find out.” He continues:

“Kurzweil compares his theory with the physical structure of the brain, hurling a huge amount of neuroanatomy at the reader, and asserting, without a lot of reflection, that it all fits his theory. A recent paper (more controversial than Kurzweil may have realized) claims that the brain is neatly organized into a kind of three-dimensional grid system. Kurzweil happily takes this as evidence that he was right all along, but the fact that the brain is organized doesn’t mean it is organized as Kurzweil suggests. We already knew that the brain is structured, but the real question is what all that structure does, in technical terms. How do the neural mechanisms in the brain map onto the brain’s cognitive mechanisms? Without an understanding of that, Kurzweil’s pointers to neuroanatomy serve more as razzle-dazzle than real evidence for his theory.”

For his part, Bailey appears to accept Kurzweil’s arguments as well as the conclusions he draws from them. He writes:

“Mr. Kurzweil thinks that we will not only augment our own minds but also create conscious, independent artificial intelligences. We will know that they are conscious because they will tell us so in a convincing way. And the super-intelligent machines we create will not replace us. ‘This is not an alien invasion from Mars—we are creating these tools to make ourselves smarter,’ he says. ‘We build these tools to extend our own reach.’ Just how far will our reach extend? Exponentially expanding intelligence, Mr. Kurzweil says, will quickly solve minor problems like war, death and material scarcity and then head out to colonize the rest of the universe. He modestly concludes: ‘Waking up the universe, and then intelligently deciding its fate by infusing it with our human intelligence in its nonbiological form, is our destiny.’ Sounds good to me.”

Well it doesn’t sound good to Marcus. He writes, “The deepest problem is that Kurzweil wants badly to provide a theory of the mind and not just the brain. Of course, the mind is a product of the brain, as Kurzweil well knows, but any theory that seriously engages with what the mind is has to reckon with human psychology—with human behavior and the mental operations that underlie it. Here, Kurzweil seems completely out of his depth.” He explains:

“Not a single cognitive psychologist or study is referred to, and he scarcely engages the phenomena that make the human mind so distinctive. There’s no mention, for example, of Daniel Kahneman’s Nobel Prize winning work on human irrationality, Chomsky’s arguments about innate knowledge that sparked the cognitive revolution, or Elizabeth Spelke’s work on cognitive development demonstrating the highly nuanced structure that is present within the mind even from an extremely early age. Similarly absent is any reference to the vast literature on anthropology, and what is and isn’t culturally universal.”

Marcus concludes that Kurzweil’s “secret of human thought” is little more than a generic theory that “has been around since the late nineteen-fifties.” He argues that there are too many questions left unanswered for anyone, including Kurzweil, to claim any certitude. He writes:

“What Kurzweil doesn’t seem to realize is that a whole slew of machines have been programmed to be hierarchical-pattern recognizers, and none of them works all that well, save for very narrow domains like postal computers that recognize digits in handwritten zip codes. This summer, Google built the largest pattern recognizer of them all, a system running on sixteen thousand processor cores that analyzed ten million YouTube videos and managed to learn, all by itself, to recognize cats and faces—which initially sounds impressive, but only until you realize that in a larger sample (of twenty thousand categories), the system’s overall score fell to a dismal 15.8 per cent. The real lesson from Google’s ‘cat detector’ is that, even with the vast expanses of data and computing power available to Google, hierarchical-pattern recognizers still stink. They cannot come close to actually understanding natural language, or anything else for which complex inference is required.”

Marcus argues, “The kind of one-size-fits-all principle of hierarchical-pattern learning that Kurzweil advocates doesn’t work on its own in artificial intelligence, and it doesn’t provide an adequate explanation of the brain, either.” Marcus accepts the fact that Kurzweil “knows artificial intelligence,” but he asserts that “Kurzweil doesn’t know neuroscience” or “understand psychology.” He concludes:

“To truly reverse-engineer the human mind, we may need a real consilience, to borrow a word from the Harvard biologist E. O. Wilson, a coming together of workers in A.I. with researchers who study the human mind from a wide range of perspectives — neuroscientists and cognitive psychologists, and maybe even artists, musicians, and writers, too. The challenge of figuring out how the mind works is too complicated for even the smartest of entrepreneurs to solve on their own.”

Having not read the book, I’ll leave final judgment of its value to those who have. My suspicion, however, is that the book will be more entertaining than enlightening if all it does is rehash ideas that have presented previously. I agree with Marcus that we have much yet to learn about the mind (as opposed to the brain). IBM just announced that it had created the world’s fastest supercomputer that contains 530 billion simulated neurons and 100 trillion simulated synapses, which equals the number of neurons and synapses in the human brain. That computer might help determine whether or not Kurzweil is correct — but that’s topic for a later date.