May 8, 1998
Steve Allen once created a short-lived television series called Meeting of the Minds. A deadpan Allen would introduce four guests, in period dress, drawn mostly though not exclusively from the ranks of the intellectual Immortals. Marx, Spinoza, Leonardo, and Marie Antoinette were a typical quartet; Poe, Jefferson, Aristotle, and Virginia Woolf were another. In all -- well, not all -- seriousness, the five would instantiate for half an hour the Great Conversation that is Culture. I was generally on the floor after ten minutes.
John Casti is not, like Steve Allen, a comic genius, just a veteran science writer with a streak of whimsy. But The Cambridge Quintet is an inspired conception. One rainy night in 1947, in his (and Charles Darwin’s) former rooms at Cambridge University, C. P. Snow, novelist and science advisor to the British government and soon to be famous for his pamphlet, The Two Cultures, has arranged a meeting of minds. The government wants to know whether there’s anything to the talk just then beginning to be heard about the possibility of “thinking machines.” So Snow has invited J. B. S. Haldane, Erwin Schrodinger, Alan Turing, and Ludwig Wittgenstein to dinner. Over five courses (and chapters), accompanied by sherry, Montrachet, Burgundy, and cognac, the conversation unfolds.
The four guests are, of course, bona fide Immortals. Haldane was an early and pioneering population geneticist, a prolific popularizer, and a leading (though eventually disillusioned) British Communist. Schrodinger was one of the founders of quantum mechanics, a dabbler in Eastern metaphysics, and author of the now-classic little book What Is Life? Turing, as much as anyone, launched the computer age with a series of revolutionary mathematical papers. (He also, as much as anyone, defeated the Nazis by breaking their previously impenetrable military-communications code, “Enigma.”) Wittgenstein was not only one of the most influential thinkers of the twentieth century but also one of its most influential personalities: ascetic, mystical, a kind of philosopher-saint.
What started the rumors that had so intrigued the British government was the invention of the “Turing machine.” This was not a concrete object, with bells, whistles, and wires, but rather the idea of an object with a certain structure and certain abilities. A Turing machine is something that can take in a symbol and perform a specified sequence of operations (called a “program”) on it, thereby transforming it into another symbol, which it prints out. Logics, grammars, and mathematical functions are also ways of processing symbols. Turing showed that symbol-processing has a general structure (called “computation”), which can be embodied physically, in a machine.
But, Snow asks, is computation the same as thought? Our dinner guests are off and running. In one of several attempts at explaining the implications of his mathematical findings (like Godel’s “uncertainty” principle and Schrodinger’s “indeterminacy” principle, Turing’s “computability” theorem is so abstract and general that it has myriad implications and is susceptible of many different formulations), Turing remarks that “anything at all that can be thought of as following a set of rules ... can be calculated in a similar step-by-step fashion by this kind of machine.”
This pushes the argument back a stage. Does the brain “follow rules”? Can its operations be formalized, or made explicit? What is this thing called thought? Apparently, it’s a conversation among neurons. The brain and nervous system are made up of roughly ten billion neurons, or nodes, connected by wire-like axons and dendrites, which conduct electrical impulses to and from. “Something like a giant telephone switching network,” as Haldane puts it. The neurons have a “threshhold of excitation”: if the incoming electrical impulses are strong enough, they fire off impulses in turn, which are carried to other neurons. The “on-off” pattern of all the neurons (or some subset of them) at any one time is a mental state -- in effect, a thought.
All this is strikingly analogous -- in some ways, at least -- to how a computer works. A neuron is like a computer “bit,” or storage location; “on” and “off” correspond to the 0s and 1s of the binary system; stimuli to the nervous system are the equivalent of inputs to the machine; the firing of neurons and their rearrangement into new patterns resembles the executing of instructions from a program and the consequent rearrangement of the stored data into new configurations.
That’s as much of the scientific logic as this reader could follow. Fortunately, there’s also plenty of non-technical talk. One of the more illuminating strands of the discussion, spread over several chapters, counterposes two famous thought-experiments: the “Turing test” and John Searle’s “Chinese Room” argument. (Searle’s argument, actually published in 1980, appears in The Cambridge Quintet, renamed and attributed to Wittgenstein, through fictional license.) Turing proposed a simple and ingenious test for machine intelligence. Put a person in one room, a computer in another, both connected by teletype to a person in a third room. The person in the third room types in questions, to which the other person and the computer type out replies. If the questioner cannot tell consistently which replies come from the other person and which from the computer, then the computer is intelligent. If it talks like a human ... .
Not necessarily, replies Searle/Wittgenstein. Put a person who knows no Chinese in a room with a very large manual or computer program that matches all possible questions in Chinese with appropriate answers in Chinese. Slips of paper with questions go under the door; appropriate answers come back. By Turing’s definition, the person in the room understands Chinese. But he doesn’t. Does this example invalidate the Turing test? Turing, with a little help from Schrodinger, mounts a defense, which persuaded me but not Wittgenstein.
Casti gives all the characters some good lines; but fundamentally the debate is between Turing and Wittgenstein, with Snow moderating and Haldane and Schrodinger in supporting roles. Dramatically and intellectually, this seems right: of the four, it’s their work that bears most directly on the question of artificial intelligence. Philosophically, Turing was a positivist, Wittgenstein an anti-positivist; polemically, Turing was diffident, Wittgenstein fierce. In The Cambridge Quintet, as in life, Turing was a troubled but sympathetic soul, an unworldly genius. Wittgenstein does not come off so well, but that is the book’s fault. His thought was obscure and his personality difficult, but he was not quite so dogmatic and ill-tempered as he appears here.
The Cambridge Quintet is an agreeable jeu d’esprit, though a bit thin. If it whets your appetite, go on to Daniel Dennett’s Brainstorms and Douglas Hofstadter’s Godel, Escher, Bach. But don’t expect to make up your mind. Important philosophical controversies are rarely resolved in the same terms in which they are posed. Instead we find a new vocabulary, or let the subject drop for a few decades (or centuries) and come back to it in a new form. In two hundred years, most creative work in science and mathematics will undoubtedly be done by electronic entities. But they won’t be much good at poetry or fiction; they won’t be witty or passionate; there won’t be a plasma Proust or a silicon-based Lawrence. Will they be intelligent? We’ll answer that question then by means of distinctions we don’t have a glimmer of now. That’s always how it happens.