California Expert Software

Introduction
B+ The Turing Test Verbal Behavior as the Hallmark of Intelligence Stuart M. Shieber, ed. London: MIT Press, 2004

A few Saturdays ago, as I left the house, I noticed a large, fat crow standing on the curb. This seemed a little unusual, as most crows are out in the fields cleaning up harvest leftovers at this time of year. When I took the County road to a shopping center the next day, there were flocks of crows scowering the fields. The fat crow on our curb wasn't one of them; instead, it was part of the flock that winters on the grounds of our condominium complex. They survive on acorns, olives, oranges, lemons and a variety of other berries and foods easily available here. They have learned to put human inventions to work for them, reducing the energy required to obtain food. Acorns are especially nutritious, and are available freshly opened by lawn mowers and foot traffic. The fat crow and many others of his tribe have also learned how to use cars as nut crackers. They place acorns on the road, exactly where cars are most likely to crush them. They have learned where that place is by experience, by observing the traffic. I think our crows are pretty smart.

Alan M. Turing proposed his Imitation Game, what we call the Turing Test, in his 1950 paper, "Computing Machinery and Intelligence," originally published in Mind. Shieber has done us the favor of reprinting this hard to get paper in Chapter 4. Shieber also includes in this book several comments and notes Turing recorded before his untimely death in 1953. Turing's death was probably a suicide after his arrest for sodomy by a bigoted British government. But Turing lives long and prospers, not just because he invented computer science, built and programmed the first electronic computers and helped to save Britain by breaking the German Enigma codes, but because he envisioned intelligent machines capable of thinking just like us.

Turing initially thought a thinking machine would be built in about 50 years, but changed his mind in later interviews, saying it would probably take 100 years. Today, 56 years after the Mind paper, we still don't have thinking machines that qualify as humans, but we do have computers galore that are far more capable than any writers in this book imagined. In inventing modern computer science, Turing created the world which is ours almost everywhere. Unfortunately, bigoted governments continued in operation through it all, exemplied lately by a United States government run by those who believe machines are soulless, thus cannot think. Homophobia is probably a factor in Turing's lack of recognition in the world at large, although he is well known in academia. This same schizophrenia applies to thinking machines, although in reverse. It appears the general public can be easily convinced machines think, as in R2D2 and his clones, whereas professors remain skeptical.

What is a Mind? What is Consciousness? How have we Homo sapiens arrived where we are? Are we really intelligent? Are we capable of voluntary choice? And what is "voluntary choice?" I believe these are not trivial questions, and, further, that no one (including myself) has yet answered them in an entirely satisfactory manner. This book, The Turing Test, would probably be of archival interest only, had all those questions already been answered. Instead, it is still relevant to the on-going discussion about intelligent creatures.

One of the persisting confusions about whether a humanly made machine can be intelligent occurs when computing is conflated with free will. Turing discusses this problem (p. 114 ff), but doesn't give a credible answer. One of the objections to machine intelligence is just that it is programmed; i.e., machines only do what they are told. In order to be intelligent, they would have to exhibit some form of free will. Turing suggests, but leaves aside, the proposal so many others have made: free will is emulated by introducing random elements into the computations. But randomness is not the same thing as free will; at least not in the way free will applies to human beings. We think of free will as the making of choices between alternatives, some reasonable and some not. Our free will is not the luck of a random draw. Turing has faith that many years of programming will eventually produce a machine that "thinks" in a manner similar to humans, which is the reason he proposed the Turing Test. (p. 116)

Oddly enough, Turing denigrated efforts to build machines that have the look and feel of humans. (p. 116) Fifty years later, that is exactly the strategy of SONY and other corporations in building robotic toys and assistants. It is what works in Sci-Fi movies, where the machines are often recognizably intelligent. In the HBO movie production, ISLAND, clones of humans are grown as spare parts for wealthy patrons, but they eventually turn into intelligent creatures. The two important assumptions of that movie are (1) any organism with something like a human brain can be intelligent and (2) human brains can be programmed. An unexpected consequence of having a human brain, and the pathos of the movie, is either that the programming is inadequate or that it is overcome. Being able to divest oneself of programming is a supremely human ability; it is the ability to invent oneself. Machines or clones capable of rising above programming are probably sufficiently intelligent to pass a Turing Test. For that reason, I think Turing erred in minimizing appearance.

Richard Purtill has some of the same thoughts in his chapter, "Beating the Imitation Game." He writes, "Thus it seems that Turing's "imitation game" is just an interesting piece of science fiction." (p. 169) He points out that computer "... output is totally determined by the program and the input." (p. 170) He concludes, "As opposed to this, hardly any behavior of thinking beings can be shown to be completely determined by "programming" (i.e. teaching or conditioning) plus external "inputs" (i.e. experience of various kinds)." (p. 170) So, in Purtill's mind, there is a huge barrier between machines and thinking, apparently the barrier between mechanical and the organic.

In Robert M. French's paper, "Subcognition and the Limits of the Turing Test," we find another criticism of Turing's Imitation Game. "We will see that the Turing Test could be passed only by things that have experienced the world as we have experienced it, and this leads to the central point of the present paper, namely, that the Test provides a guarantee not of intelligence but of culturally-oriented human intelligence." (p. 184) I think this criticism is essentially correct, but it does not lead where French thinks. In a series of arguments, he shows "... the physical level is not dissociable from the cognitive level." (p. 194) He concludes, "... any attempt to "fix" the Turing Test so that it could test for intelligence in general and not just human intelligence is doomed to failure because of the completely interwoven and interdependent nature of the human physical, subcognitive and cognitive levels." I agree with that statement, but, again, it does not lead to French's conclusion. What seems to me the case is that each creature probably has intelligence unique to itself. That characterization applies right down to specific human beings. So, French's arguments only beg the question, how do we recognize intelligence in others (of our own or any other) species? How do I know the crows outside my house are smart?

Perhaps the most acidic paper in this book is John R. Searle's "Minds, Brains and Programs." He states views later expounded in his Berkeley course, and detailed in MIND (Oxford: Oxford University Press, 2004). [Review pending - ed.] He rejects the claims of "Strong AI" [his term], "... that the programmed computer understands the stories and that the program in some sense explains human understanding." (p. 204) His invention for the purpose of refuting Strong AI is the Chinese Room. He supposes that he is locked in a room and made to answer in Chinese questions posed in Chinese, but, of course, he is only an English speaker. Suppose, further, that Prof. Searle succeeds in convincing his Chinese inquisitors as to the wisdom of his answers, which he does not understand at all. He just receives squiggles (input) and sends other squiggles back (output). He asks, "... what is it that I have in the case of the English sentences that I do not have in the case of the Chinese sentences?

This begins an extensive analysis of the Chinese Room. The first rebuttal to Searle's problem is "The Systems Reply (Berkeley)," which is essentially that mind is everywhere. (p. 209 ff.) In other words, the intelligence inheres in The System, not in how it is implemented. Second is "The Robot Reply (Yale)," which fails to explain what is "understanding" according to Searle. (pp. 211-212) The lack of understanding is a fault in "The Brain Simulator Reply (Berkeley and M.I.T)." He asks, "... where is the understanding in this system? It takes Chinese as input, it stimulates the formal structure of the synapses of the Chinese brain, and it gives Chinese as output. But the man certainly doesn't understand Chinese ... The problem with the brain simulator is that it is simulating the wrong things about the brain. ... it won't have simulated what matters about the brain, namely its causal properties, its ability to produce intentional states [consciousness - ed.]." (p. 213)

Searle makes an effort to rebut "The Combination Reply (Berkeley and Stanford)" which appears to fix the problems of the previous defenses of Strong AI. "Imagine a robot with a brain-shaped computer lodged in its cranial cavity, imagine the computer programmed with all the synapses of a human brain, imagine the whole behavior of the robot is indistinguishable from human behavior, and now think of the whole thing as a unified system and not just as a computer with inputs and outputs. Surely in such a case we would have to ascribe intentionality to the system." (p. 213) Searle refutes the combination defense of Strong AI in this story:

"To see this point, contrast this case with cases in which we find it completely natural to ascribe intentionality to members of certain other primate species such as apes and monkeys and to domestic animals such as dogs. The reasons we find it natural are, roughly, two: we can't make sense of the animal's behavior without the ascription of intentionality and we can see that the beasts are made of similar stuff to ourselves ... We would certainly make similar assumptions about the robot unless we had some reason not to, but as soon as we knew that the behavior was the result of a formal program, and that the actual causal properties of the physical substance were irrelevant we would abandon the assumption of intentionality."

While that is an interesting argument, it cuts two ways. For, if we suppose the human-like robot is, in fact, a human being, we might abandon the hypothesis of self-conscious, self-directing people. It is only because Searle and others prefer to believe of themselves and others that they are not robots, that this argument has any effect. When the prejudice favoring our own exalted status is removed, it can admitted humans actually behave much like robots; i.e., they are zombies. I think what Searle's argument comes down to is 'appearances can be deceiving.'

The other take on Searle's argument is just that it is all a matter of definition. Once we define a machine run by a program as lacking intentionality, it is always a robot, and at best a zombie. But, then, what is not programmed? In what sense are organic creatures different from man-made machines, especially the Frankensteins of the so-called Combination Reply. I find Searle's argument difficult to accept, unless one posits some form of soul or vitalism, or unless he wants to reduce everything to just a way of talking.

Shieber identifies Searle and others as "... the essentialists whom think that there is something intrinisc about people's brains that make them intentional, thinking beings .." (p. 268). That thought introduces an article by Prof. Daniel C. Dennet, "Can Machines Think?," defending Turing's thesis. Dennett's summary argument is this:

"If you look at a computer ... if you open up the box and look inside and see those chips, you say, "No way could that be conscious. No way could that be self-conscious." But the same thing is true if you take the top off somebody's skull and look at the gray matter pulsing away in there. You think, "That is conscious? No way could that lump of stuff be conscious." (p. 290)

Dennett concludes, "If you want to get a grasp of how a computer could be conscious, it's no more difficult in the end than getting a grasp of how a brain could be conscious." (p. 291)

In the light of recent history, the two most important philosophers cited in this volume are Searle and Dennett. Turing's Test is still alive and well, despite Searle's damning criticisms. Dennett's relative criteria about the difficulty of the problem still stand. Whether machines can think is still an outstanding question.

Here, I add this confession for those who didn't know it: with Dennett, I believe machines can think. I also believe crows do some thinking, and so do some humans as well. I also believe Searle's criticisms reveal the problem is to some extent definitional, but my view is possible because I am not an essentialist.

I intend to continue discussing the issues Turing put before us over a half century ago. Shieber's compendium is a very useful text in summarizing Turing's Imitation Game and the controversy it continues to generate.