The Physicist Against the Hackers : THE EMPEROR’S NEW MIND: On Computers, Minds, and the Laws of Physics <i> by Roger Penrose (Oxford University Press: $24.95; 428 pp.) </i>
Could a computer ever think? Be intelligent? Have a mind? Be creative? Be conscious? Or is there something special about humans that endows us with these wonderful properties but, being absent in current computers, dooms them to a dark and mindless existence?
Proponents of “strong artificial intelligence” (AI) claim that intelligence and consciousness arise somehow from the way in which a device computes--described by its set of rules, its algorithms . Thus consciousness can be evoked not only in biological brains but also in properly programmed computers, or, for that matter, in an astoundingly complex contraptions of gears and pulleys. The task, according to such proponents, is to ascertain the proper algorithms.
Not so, claims Roger Penrose, the distinguished mathematical physicist from Oxford University; the brain is uniquely suited to harness non-algorithmic aspects of the world, and thus can be conscious. As its title suggests, “The Emperor’s New Mind” is Penrose’s attack on strong AI; it is based on recent research on quantum mechanics, quantum gravity, black holes, cosmology, fractals, quasicrystals and more, all as they relate to the question of algorithmic processing.
His general argument builds on the ideas of philosopher Karl Popper. First, Kurt Goedel’s celebrated theorem demonstrates that there are mathematical truths that are forever beyond proofs using merely the algorithms of mathematics itself--which is all that a standard computer could employ. Next, conscious beings seem to be able at times to recognize truths directly, apparently without deriving them by any algorithms. Thus, in his effort to understand consciousness, Penrose seeks properties of the physical world involving non-algorithmic behavior, aspects which biological brains might exploit. But are there any properties of the physical world such that their behavior can not be captured in a computer program?
For Penrose, astounding complexity of behavior and chaos such as the billowing dust from a volcano will not suffice, since this could in principle be simulated in a very large computer. Nor will quantum mechanics--the branch of physics typically treating very small or very fast processes such as the decay of a radioactive atom--suffice, despite its “magical” aspects such as the wave-particle duality, “collapse of the wave function,” and so on. The problem here is that, as quantum mechanics is currently understood, random probabilities underlie all phenomena--hardly what Penrose would like in a being that has will, consciousness, intelligence and ethics.
Penrose’s central hypothesis is that a current dilemma in physics--how to reconcile quantum mechanics with Einstein’s relativity--may hold the clue. If a so-called “correct quantum theory of gravity” ever is devised, perhaps it will be non-algorithmic.
What could be more intriguing--a book that relates forefront physics research with the age-old philosophical question: “What is consciousness?”
But, frankly, the book is disappointing--and at times frustrating--despite the clarity of the writing and the manifestly interesting topic. Time and time again Penrose defers objections or alternate interpretations, breezing along with his line of thought; when he finally has to meet these objections--often quite late in the book--many of his “explanations” collapse.
Here is just one example. The crushing disappointment concerns his central hypothesis that the key to having a brain function non-algorithmically is to exploit phenomena of an as yet undiscovered “correct quantum gravity,” for which standard quantum mechanics is but an approximation. He is forced to postulate special nerve cells in the brain-- as yet not recognized --that operate on fundamentally quantum gravitational processes in a manner as yet not even postulated . After about 100 pages related to such wild speculations, he then admits that the brain is too “hot” for the scheme to work anyway; the random motion of its components would swamp any quantum gravity effect. He concludes this analysis somewhat despondently: “Perhaps we are doomed to be computers after all!”
After Penrose paints himself into an unpleasant corner this way, he tries to paint a doorway out. It simply doesn’t work. He turns in the last chapter away from his field of expertise--physics--and into territories where he is, frankly, an unreliable guide; for instance, psychology, cognitive science and the evolution of consciousness. Some of his analyses here are mere recapitulations of ideas presented elsewhere.
Conversely, there are issues that have preoccupied other thinkers on these subjects--most notably the importance of levels of description, and flexibility of response--that get only cursory treatment by Penrose.
Although the current status of computers and algorithms is not central to Penrose’s argument--he is talking about computers “in principle,” after all--what little he does describe is surprisingly out of date; for instance, computer “psychoanalyst” programs, now nearly two decades old. When he mentions the current status of chess-playing computers, he can tell us only the level (grand master) at which they perform, giving us no insight into whether any new ideas were necessary for such improved performance. Since he feels that the deduction of truths outside a mathematical system is unique to humans, it would have been interesting to see what he would have written about recent computer programs that have found mathematical proofs not devised by humans.
His intuitions concerning computers seem to be based on limited experience in front of a video screen with canned programs, rather than with flexible and adaptive computer systems that interact with the environment. In the last few years, there have been important new insights and successes incorporating the interactive and flexible function of the type in biological brains into computer programs--variously known as neural networks, parallel distributed processing, or connectionism--that have led even some stalwart opponents of strong AI to now embrace the concept of computer mind.
Surely no one can fault Penrose for failing to solve the mind-body problem; but his book, given its frequent reliance on narrow intuition and the unpersuasiveness of its central, novel hypothesis, should be viewed with skepticism by the lay reader and only approached after checking on any of a number of good books (on both sides of the strong AI debate): “Brainstorms” (D. Dennett), “Godel, Escher, Bach” (D. Hofstadter), “The Self and Its Brain” (K. Popper and J. Eccles), “Neurophilosophy” (P. S. Churchland) or “Minds, Brains and Science” (J. Searle).
The greatest value in “The Emperor’s New Mind” is its clear descriptions of difficult but fascinating physics, material that will interest the lay reader even though it may have little direct relevance to the strong AI debate. The analyses of fundamental issues in quantum mechanics such as the “Schrodinger’s cat” conundrum, the origin of the “direction” of time (past to future), and the process surrounding the creation and destruction of black holes all are presented very well. The mathematics accompanying the explanations is illuminating, though rarely essential to an understanding.
The profound mind-body problem and the strong AI debate come to all of us at one time or another in our reflective moments; it motivates many of us working in the field. Although we are surely a long way from demonstrating or refuting the strong AI hypothesis (indeed it might never be demonstrated to everyone’s satisfaction), any important advances will require insights from a variety of disciplines: philosophy, computer science, psychology, mathematics, biology, and so on. The greatest benefit Penrose’s book may have is to help attract more physicists into the debate.
We need all the help we can get.
