stty consulting › our future
Review of Penrose, The Emperor's New Mind
Review of The Emperor's New Mind Concerning Computers, Minds, and the Laws of Physics.The Times Literary Supplement. Oxford University Press
The Times Literary Supplement, September 29-October 5, 1989. [review of] Roger Penrose, The Emperor's New Mind: Concerning Computers, Minds, and the Laws of Physics, Oxford Univ. Press, 1989.
The idea that a computer could be conscious--or equivalently, that human consciousness is the effect of some complex computation mechanically performed by our brains--strikes some scientists and philosophers as a beautiful idea. They find it initially surprising and unsettling, as all beautiful ideas are, but the inevitable culmination of the scientific advances that have gradually demystified and unified the material world. The ideologues of Artificial Intelligence (AI) have been its most articulate supporters. To others, this idea is deeply repellent: philistine, reductionistic (in some bad sense), as incredible as it is offensive. John Searle's attack on "strong AI" is the best known expression of this view, but others in the same camp, liking Searle's destination better than his route, would dearly love to see a principled, scientific argument showing that strong AI is impossible. Roger Penrose has set out to provide just such an argument.
It is a huge project. In order to build his case, Penrose must lead the reader through detailed discussions of many topics in mathematics (Turing machines and computability theory, complex numbers, the Mandelbrot set, Gödel's theorem, recursive function theory, complexity theory, Platonism versus intuitionism), classical Einsteinian physics and quantum physics, cosmology, and, of course, neuroscience. Most of these topics have been given excellent popular presentations in recent years--in Hofstadter's Gödel Escher Bach (Basic Books, 1979), Hawking's A Brief History of Time (Bantam, 1988), Gleick's Chaos: Making a New Science (Penguin, 1987)--but Penrose believes that he must go over this material again in his own way, digging deeper, explaining in more detail. The result is bracing reading, to say the least, and the topics for hundreds of pages on end apparently have nothing to do with the mind at all.
The inevitable first impression, then, is that the book is the ultimate academic shaggy dog story, a tale whose fascinating digressions outweigh the punch line by a large factor. Why does Penrose do it? Is there no swifter, more accessible route to his conclusion? No. Penrose sees that he has no hope of overthrowing the case for strong AI unless he can dislodge one of the most imperturbable objects in the intellectual universe: something I will call the Cathedral of Science.
The Cathedral of Science is the highly structured, beautifully articulated amalgam of "what everyone should know" about science, crowned by the inscrutable but talismanic formula, "e=mc2.". Its facade, visible to the general public, is popular lore: familiar and decorative items of information and misinformation about the Galilean physics of everyday objects, cartoon-style renderings of black holes and language-using chimps, and pockmarked with such tidbits as "you only use five percent of your brain" and "no two snowflakes are alike." Items in this layer are easily replaced or swept away, but underneath it lies the scientists' (and philosophers') much denser version of the same material, created largely of the remembered oversimplifications of university-level textbooks, supplemented by New Scientist and Scientific American articles, and such other high-quality interdisciplinary communications as the books just mentioned. This material forms the communally shared understanding on which everyone relies while working on their more particular specialities. Aside from a few brilliant polymaths, the neuroanatomist, the biochemist, the experimental psychologist and the philosopher of mind have roughly the same workaday understanding of quantum mechanics, entropy, and computability, for instance, and this understanding gives them sufficient reason to believe that they needn't understand these topics any better in order to do their work. The Cathedral's architecture is the familiar hierarchy of mechanistic materialism: living bodies are self-preserving, self-replicating machines made out of cells made out of molecules made out of atoms--with some weird quantum physics isolated (one hopes) in the cellar. No church has ever enjoyed a more entrenched or authoritative orthodoxy, an empire that expands with daily discoveries and protects itself from swift change by the distributed, mutual myopia of its adherents. Its heresies (ESP, creationism, vitalism) are readily identified and deplored in unison; its conservatism is hailed by almost all who participate in it, and for good reason.
It is Penrose's immense task to restructure our vision of the Cathedral of Science, shaking our conviction that it is largely settled and safe and familiar (except, of course, for that baffling business about quantum physics). His task is made all the more intricate by his recognition that most of the Cathedral is sound. He is a revolutionary, but no bomb-throwing nihilist. Like Archimedes, he needs a place to stand if he is to move the world, so he introduces a new taxonomy of theories in science, SUPERB, USEFUL and TENTATIVE, to distinguish what is inviolable and must somehow survive any revolution, from what might be replaced or abandoned. Euclidean geometry, Galilean dynamics, Maxwell's equations, Einstein's special and general relativity theories, quantum physics and quantum electrodynamics are all SUPERB, but even these must be put into a new alignment if we follow Penrose.
Briefly, here is the path of Penrose's proposed revolution. If the brain is a computer, its powers are circumscribed by the limits on all computation uncovered by Turing and Gödel. Turing showed that each possible mechanical computation can be precisely specified by a recipe consisting of a sequence of dead-simple mechanical steps. Such a recipe is called an algorithm; all computer programmes are algorithms. Gödel's Theorem showed that no algorithm for proving mathematical truths can prove them all. Doesn't Gödel's Theorem establish that there are tasks "we" (mathematicians, in any event) can perform that are beyond the capabilities of any machine? The idea that the human race can be saved from machinehood by riding on the coattails of those clever enough to understand Gödel's Theorem is well-explored territory, and the received wisdom is that all the previous arguments for this conclusion have been roundly defeated, so if Penrose is to get his needed premise here, he must find a new wrinkle. The standard Cathedral vision is that Gödel's Theorem proves that there is just some single arcane truth of number theory (a machine's Gödel sentence) that is beyond all mechanical computation (by that machine), and Penrose's detailed exposition of the wealth of non-computable but knowable results replaces that vision with an appreciation of the depth and importance of the realm of non-computable mathematics, certainly a domain that is eminently accessible to human mathematicians relying on "insight". Moreover, the results of complexity theory show that there are many officially computable results that are not practically computable--the algorithms that are guaranteed to yield the answer would take billions of years to run on the fastest conceivable computers. How, then, do "we" arrive at solutions to these problems? Penrose proposes that there is a "theoretical possibility that a quantum physical device may be able to improve on a Turing machine." (p.146)
This leads then to a solid review of classical (non-quantum, but relativistic) physics, packed with novel perspectives and designed to impress us that "we should not be too complacent that the pictures that we have formed at any one time are not to be overturned by some later and deeper view." (p.217) With our minds thus stretched open, we plunge into "quantum magic and quantum mystery," and are led yet one more time through the two-slit experiment, Heisenberg's Uncertainty Principle, the collapse of the wavefunction, the Einstein Podolsky Rosen paradox and Schrödinger's notorious cat--in more detail than I have heretofore encountered in a popular book. Here the upshot is more radical: Penrose doubts that the puzzles of quantum theory and its relation to classical theory will succumb to any tidy, local resolution, and, like Einstein, he resists the standard "anti-realist" interpretations favored by most physicists. After a further chapter laying groundwork in cosmology on the flow of time and the curious status of the second law of thermodynamics, we are ready for the suggestion that if a unified theory is to be found, it will have to be a theory of "quantum gravity," requiring "a change in the very framework of the quantum theory." (p.348) At this time Penrose can present only speculations about the "germ" of such a theory, which is not yet, in his own terms, even TENTATIVE, so he has to settle for some gestures in the direction he feels the revolution will take.