I was three-fourths of the way through this book before I began to suspect that it wasn't headed where I thought. I had been carried along by Penrose's clear writing and obvious mastery of a wide variety of very interesting subjects, fully expecting, on the basis of the introduction and cover copy, Penrose to be leading up to another unconvincing argument for why the mind cannot be a computer. I've read several such arguments, such as John Searle's (in)famous "Chinese Room" argument, which runs something like this:
Let's say we have a computer programmed to pass the Turing test, in this limited sense: We feed it simple stories, like "A man went into a restaurant and ordered a hamburger; when it arrived, it was burnt to a crisp, and he stormed out of the restaurant without leaving a tip." Then we ask it yes or no questions like "Did the man eat the hamburger?" And let's say the stories and the questions are in Chinese.
In another room is the philosopher John Searle, who knows not a word of Chinese. We shoot the stories and questions in to him as to the computer, and he works through the same algorithm, coming up with the same yes or no answers. Because he knows no Chinese, he obviously doesn't understand the stories, nor the questions to which he is providing correct answers. If he doesn't understand the content he's operating on mechanically, then it's surely silly to suggest that the computer -- performing the same operations -- does.
Penrose finds Searle's argument compelling but not entirely convincing. I'm probably not doing the argument justice, because as stated here it seems foolish to me. Whatever understanding might turn out to be, Searle has set up a problem domain in which it is not required, and then faults the computer for not employing it. That hardly seems fair.
But Searle's argument is not Penrose's argument. Penrose does lead up to an argument that the mind cannot be a computer, but by the time I was three-fourths of the way through the book, it was apparent that he had a new approach, and that it is not one that can easily be dismissed.
A priori, one would not expect to be able easily to dismiss a Roger Penrose argument. Penrose is a mathematician who knows a lot about modern physics and has won a lot of awards, including the 1988 Wolf Prize, which he shared with Stephen Hawking for their joint contribution to our understanding of the universe.
In building his case, Penrose touches on the computer science and mathematics themes of the Turing test, the Church-Turing thesis, Hilbert's problem, Godel's theorem, complexity theory, and the Mandelbrot set; physics topics such as special and general relativity, quantum theory, cosmology, entropy, the big bang, and quantum gravity; brain research issues like split-brain studies and brain plasticity work; and philosophical issues like quantum vs. Platonic views of reality, and the distinction between computability and determinism. Because he really uses all these disciplines in developing his argument, only someone with a firm grasp on all these subjects could properly evaluate the argument. That's not me; but the group of people capable of evaluating the argument may be limited to Penrose.
Penrose's argument, though, is easy to state. To understand something, according to Penrose, is to make direct contact with Truth.
Some readers may feel that this raises some questions. Penrose actually invites us to ask, "What is truth?" Penrose's answer is a return to a Platonic view of a mathematically pure reality "out there" somewhere. This is not a trendy view. Nick Herbert's book Quantum Reality well characterizes the views on reality held by or entertained by modern physicists, and none of them take this naive Platonic approach.
Penrose realizes that this view is not consistent with quantum physics, but it does seem consistent with classical physics, which seems to describe the world of everyday experience. Penrose agrees with Einstein and others that quantum physics must be incomplete, and that there is some generalization waiting to be discovered which will tie it together with classical physics. Discovered, not invented. Penrose really thinks there is a reality out there to be discovered. He cites the Mandelbrot set as an example of the mathematical reality that is "out there."
If there is an objective reality, then truth is simply the accord of model with reality, which is what we all thought it was all along, right? Well, we did if we weren't well versed in quantum views of reality. This naive view of truth gives Penrose a way to deal with consciousness. Consciousness is contact with reality.
This is where Penrose pins his argument that the mind cannot be a computer. One of the things that we attribute to the mind and don't attribute to any existing computer is consciousness. Nor does anyone have any notion how a computer might attain consciousness. This is not surprising, since we have no idea how humans attain it, either, or any consensus on what consciousness is.
Penrose offers a common sense definition of consciousness, and argues that a computer can't have consciousness so defined. To do so, he has to provide a limiting definition of what a computer can be, but not too restrictive: Basically, I think, he's deriving a definition of computer from the accepted definition of the adjective computable.
Along the way, Penrose presents the Godel theorem argument. We can know that an assertion in a system is true even though we cannot derive it via algorithmic procedures of the system. But he is not simply reprising the unconvincing argument for the superiority of minds over computers: that the mind can somehow step outside the system to which the computer is algorithmically bound. That argument has always seemed to me to be wishful thinking: Computers can operate at different symbolic levels; and the fact that human minds can step outside certain systems to examine problems at a higher level, does not mean that the mind can leap to any level. Nothing in our ability to see an individual problem from the outside, to bring insight to it, constitutes a proof that thought is not governed by complex algorithms.
But Penrose has something different in mind. It isn't just to a higher model that we jump, but to the thing modeled. We recognize the truth by comparing the model with reality. This must be the case, Penrose argues, because no undirected algorithm can recognize truth. The algorithmic method can lead to falsehood as easily as to truth.
Penrose doesn't say where this business of the mind touching truth occurs. He presumably believes that we'll need this new model of physics, the one that transcends quantum mechanics, to understand the connection of mind and reality. He has some thoughts about such a model, this being in his line of work, after all. Some hints, he suspects, are hidden in our perception of time.
The arrow of time, which pierces us all eventually, is not, apparently, a fact of nature, but a fact of perception. Why are we cursed to this treadmill? Penrose doesn't know, but he presents some research from physiological psychology that was new to me, showing that, while our reflexes act in fraction-of-a-second times, it takes something like a second or two to decide to act and follow through, even for something as simple as making a fist.
It's common experience that reflexes are faster than conscious decisions, but these numbers seem obviously wrong. You can do the experiment yourself, timing how long it takes you to decide to clench your fist and then to do it. Not two seconds, surely.
But the experiments were apparently well designed and controlled. What's going on here? Penrose argues that, if the results are accurate, our perception of time is even more out of step with reality than we thought. Are there some stages of conscious decision making that are unconscious that we sort of sleep through? Or do our minds simply attach a perception of near-instantaneousness to a phenomenon that actually takes a second or two?
How all this ties into the interrelationships among minds, computers, and the laws of physics is something even Penrose isn't quite sure about. The book shows what Penrose is thinking, and not all the thoughts have gelled. But the least certain of his arguments may be the most important.
Copyright © 1991, Dr. Dobb's Journal