Beating the AI Drum

Having received I Am a Strange Loop for my birthday, I have been reading it with delight. I hope to finish it either today or tomorrow, and as it deserves a full review, I expect to be giving it one shortly. However, one little (or is it?) point of disagreement has popped up.

Douglas Hofstadter, author of that book and (more famously) of Gödel, Escher, Bach: An Eternal Golden Braid, speaks at some length about the topic of artificial intelligence. In Herr Hofstadter's view, there is no metaphysical bar to machines' being able to think, though machine thought may not be precisely similar to human thought. In any case, anyone who has read his books gets the sense that he believes AI to be inevitable.

Compare John Searle; Hofstadter does it, so we might as well do it, too. Searle's objection to AI is that machines cannot possibly think like living organisms think. I think Searle may have a point - the only "machines" in our experience who are capable of thought are biologically constituted. It is not a necessary inference that only biological organisms have the physical structure needed to think; then again, the inference is appealing. It is possible that only an organism could possibly think, and our experience of things that do think is a reflection of a general prohibition on machine intelligence.

Hofstadter attacks that idea and attacks Searle personally, rather viciously, in his latest book. This is what I mean:

[John Searle] has gotten a lot of mileage out of the fact that a Turing machine is abstract machine, and therefore could, in principle, be built out of any materials whatsoever. In a ploy that, in my opinion, should fool only third-graders but that unfortunately takes in great multitudes of his professional colleagues, he pokes merciless fun at the idea that thinking could ever be implemented in a system made of such far-fetched physical substrates as toilet paper and pebbles...

What Hofstadter is saying is that a Turing machine could be constructed of anything; Searle recognizes that toilet paper and pebbles could provide the tape and on/off indicator for cells of a Turing machine; and Searle imputes to AI fanatics the absurd claim that toilet paper can (or could) think.

Elsewhere:

Such successes give the lie to the tired dogma endlessly repeated by John Searle that computers are forever doomed to mere "simulation" of the processes of life.

Well, all right; but who's the dogmatist here? Here is one dogma: "Computers can never think in the way that human beings think, because the components of computers are not suitable to thought." But here's another: "Computers can think just like humans, because the components of a computer are not different, in any way essential for thought, from the components of the human brain." Reflect on that a bit - it's just as dogmatic to assume that the cells of the human brain are irrelevant to thought as to say that they are necessary for thought. Perhaps computers can think, but Douglas Hofstadter and other AI proponents have just as strong a burden as their opponents to prove their point. People assume that the human brain is a type of computer. Then a circuit-based computer, obviously being a computer as well, ought to be able to think in a substantially similar way to the human computer. Sure; if the brain is a computer. Let's see the proof of that dogma first.

Searle may or may not be correct, but he has an interesting point. Organic compounds may have a special suitability to being the physical substrate of the abstract activity of thought. If so, reproducing that activity in a different medium may be impossible, or extremely difficult. We should not blithely expect it to happen. Further, given the demands of evolution, we should not be surprised that the configuration of our brains is the best possible (the only possible?) way to produce thought and consciousness.