Can Computers Think?

The mind/body problem is a philosophical litmus test. Here the reigning materialist view of the world — that mat­ter and matter alone exists — clashes with common sense. How can we understand our­selves as conscious, free, ra­tional agents in a universe that science tells us consists ulti­mately of mindless, meaning­less material particles? Within cognitive science, an interdis­ciplinary research program tak­ing its inspiration from the computer to understand intel­ligence, the mind/body prob­lem takes the provocative form: Can computers think?

To answer this question, Alan Turing proposed what is now known as the Turing Test: Hide a computer in one room and a human in another and see whether a human in­terrogator can guess which is the computer on the basis of responses to any questions that might be asked. Turing confidently predicted in 1950 that “at the end of the century the use of words and general educated opinion will have altered so much that one will be able to speak of machines thinking without expecting to be contradicted.”

The history of cognitive science — its emergence with­in philosophy, psychology, computer science, linguistics, and the border disciplines of anthropology and neuroscience — is lucidly chronicled in Howard Gardner’s The Mind’s New Science. Cognitive science emerged in the 1950s and 1960s as a critical reaction to behaviorism. Noam Chomsky, whose work instigated the cognitive revolution in linguis­tics, noted that the very ex­pression “behavioral sciences” seems to confuse research method and subject matter: Calling psychology a behavior­al science is as mistaken as calling physics a science of meter readings. Gardner, him­self a respected cognitive psy­chologist, finds much to ad­mire and much to criticize, in­cluding the adequacy of com­puter modeling of the mind. Anyone who is either in­trigued by the explanations given by cognitive scientists to the mysteries of the mind or annoyed by their pretensions will find this book fascinating reading.

If the cognitive revolution was sparked by a reaction to behaviorism, it was fueled by the advent of the computer. Despite disciplinary differenc­es, the cognitive scientists share the common assumption that there are mental represen­tations and that these repre­sentations can be analyzed in­dependently of the neurophys­iology of the brain. One strain of artificial intelligence known as “strong AI” postulates that the mind is to the brain what a program is to its computer hardware: Mental representa­tions are computer programs. Just as software can run on different computers, minds can be implemented naturally within brains or “artificially” on computers.

The infiltration of AI research programs into universi­ties led to the establishment of new intellectual dominions, and the emperors of strong AI have not been modest in their proclamations. Herbert Simon of Carnegie-Mellon says we al­ready have computers that literally can think, and his col­league Alan Newell says we have “discovered” that think­ing is just a matter of symbol manipulation. Marvin Minsky of MIT claims that our minds are simply “computers made of meat,” and John McCarthy, who with Minsky coined the term “artificial intelligence,” asserts that even “machines as simple as thermostats can be said to have beliefs.” John Searle, a philosopher at Berke­ley with a reputation as an in­tellectual prizefighter, retorts: “As a philosopher, I like all these claims for a simple rea­son. Unlike most philosophical theses, they are reasonably clear, and they admit of a simple and decisive refuta­tion.”

In Minds, Brains and Sci­ence, Searle attempts to deal a knockout blow to strong AI by proposing a variant of the Tur­ing Test. Suppose you’re locked in a room and commu­nicate with an outside interro­gator by exchanging pieces of paper. The twist is that all the messages are written in Chi­nese. You, however, have no knowledge of Chinese but manage to produce answers to questions by enacting the steps of a computer program. Now, a computer program is an algo­rithm — an exact, “mechani­cal” procedure that’s so com­plete that no insight, intuition, or imagination is required. To the outside observer it may appear as if you understand Chinese. However, even though you implement the computer program and behave as if you understood Chinese, you don’t. Instantiating the computer program isn’t enough for you — or any computer — to understand Chi­nese.

While denying that com­puters can think, Searle still holds that “in one sense, we are all machines…[and] of course we can think. So, trivi­ally, there are machines that think.” Brains cause minds, says Searle, in a way analogous to how solidity at the macro level is caused by the configuration of atoms at the micro level. Searle is forced to admit, however, that his reductive materialism leaves no room for free will. Minds also contain mental states that are intentional insofar as they are about the world. But when pressed for details about this material intentionality, Searle only waves his hands at some mysterious causal property which biochemists are going to have to unravel much as they did photosynthesis.

In a recent surprise best­seller, The Emperor’s New Mind, renowned Oxford mathemati­cian and award-winning theo­retical physicist Roger Penrose takes issue with both Searle and the emperors of AI. Watching a BBC program on AI, Penrose, irritated by the exaggerated claims, was prompted to write his book. Penrose writes, “the very fact that the mind leads us to truths that are not computable convinces me that a computer can never duplicate the mind.” To substantiate his claim, Penrose takes us on a dazzling tour of such topics as the Turing Test, mathematical Platonism, quantum mechan­ics, his work on black holes with Stephen Hawking, and the structure of the brain. Although computers were used to discover the intricate fractal-like structure known as the Mandelbrot set, this set cannot technically be computed by any known algorithm; and Goedel’s famous incomplete­ness theorems have as “a clear consequence that…truth cannot be encapsulated in any formal scheme.” Like Newton and Einstein, Penrose has a profound sense of humility and awe not only toward the Platonic realm of pure mathe­matics but also toward the physical universe.

Challenging strong AI and Searle’s reductionism on an­other front, Penrose speculates that human intelligence may outstrip artificial intelligence because it exploits physics at the quantum level. It is impos­sible to summarize Penrose’s views which link relativity, quantum physics, and the phenomenon of consciousness. Readers intrigued by these possibilities should take the tour for themselves. The up­shot is that the computer mind constructed by the emperors of strong AI is about as substan­tial as the Emperor’s new clothes.

Perhaps unintentionally, Penrose also exposes the scan­tiness of the reigning material­ist worldview. The mind/body problem can only be coherent­ly formulated if we have a definite conception of body, but the language we have used to talk about minds and matter is haunted with the ghosts of philosophical theo­ries. Descartes reasoned, “I think, therefore I am,” while doubting the existence of all material bodies, and so con­cluded that he was essentially a thinking substance. But the Cartesian dualism of non-think­ing matter and nonmaterial mind only exacerbates the mind/body problem: How can an immaterial mind cause physical events in space and time?

Behaviorism, while reject­ing Cartesian dualism, never­theless assigned matter to the category of the objective and the mental to the category of the subjective, and so sought to banish subjective mental events from the domain of objective science. But a Carte­sian insight still holds: If it seems to me that I’m con­scious, then I am conscious. The existence of subjective mental states are objective facts like any others. Rejecting behaviorism, cognitivism still inherited its philosophical blind spot. Ironically, the AI literature is often filled with denunciations against some view called “dualism,” but in fact strong AI is committed to a kind of residual dualism. If algorithms can exist independently of their material support, then algorithms themselves have a kind of Platonic exist­ence. The proponents of strong AI can’t hold that algorithms only exist in the mind, for then it would be circular to explain minds in terms of al­gorithms.

The naked truth is that materialism currently lacks any clear and definite conception of body, and this absence of any demarcation robs mate­rialism of its polemical preten­sions. The Cartesians offered a fairly definite conception of body in terms of contact mechanics — billiard ball cau­sality. But Newtonian physics rejected this Cartesian concep­tion by introducing gravita­tional force as “action as a distance.” Newton himself referred to gravitational forces as “occult” and retreated to the position that his theory only gave a precise mathemati­cal description of the events of the physical world. So it was not the Cartesian theory of mind but the Cartesian theory of body that was found to be inadequate; more importantly, no definite concept of body finally emerged. If it turns out that the material world is whatever we scientifically dis­cover it to be, with whatever properties it must have to serve the purposes of explana­tion, then “materialism” is too ill-defined to preclude alterna­tive worldviews that embrace the reality of mental states, free will, consciousness, and perhaps even souls.

In the original tale it is a child who asks about the Emperor’s new clothes. These books provide us with enchanting glimpses of the mind’s new science and chal­lenge us with philosophical questions it too must now face — questions so profoundly simple that even a child might ask.