It is Alan Turing’s 100th birthday today, June 23. Earlier this week I went to a documentary about Turing at Waterloo’s Institute for Quantum Computing and then did some reading. Here are a few things I found interesting...
Early computers were referred to by many names, including radio brains, universal Turing machines, and automatic computing engines.
Turing didn’t just envision the computer; he worked on teams that built some. The public archives at alanturing.net preserve fascinating letters and memos from that work.
Turing apparently committed suicide by eating a poisoned apple. Turing’s friend Alan Garner (one of my favorite authors) wrote recently that Turing had “a fascination with Disney's Snow White and the Seven Dwarfs, especially the transformation of the Wicked Queen into the Witch. He used to go over the scene in detail, dwelling on the ambiguity of the apple, red on one side, green on the other, one of which gave death.” (link)
After being convicted of gross indecency for being gay, Turing worried that his work would be discredited. He wrote facetiously, “Turing believes machines think, Turing lies with men, Therefore machines do not think.” But that argument may also be taken to be causal, as the turmoil of his last years meant that he never published his neural net sketches of intelligent machinery or his ideas for how to program.
I don't find his famous question, Can machines think?, at all interesting. It seems ridiculous: if the Turing Test determines whether machines think, and machines fail the test until something is added to them that makes it possible for them to pass the test, then it follows that machines thought after they passed the test but did not think before they passed the test - even though they did substantially the same thing. Obviously I'm wrong: the Turing Test kicked off the study of artificial intelligence and was enormously significant. Turing's thinking about the similarities of machines and human brains seems also to be based in his understanding that human thinking, including intuition and originality, are computable processes that can be replicated by machines. And of course it's increasingly possible; we could program computers to have emotional responses, lizard brain responses, a collective unconscious, and fallibility (along with things like heuristics and fuzzy logic).
So much of the writing about Turing dwells on the salacious and pathetic. It's his birthday and it seems a day to celebrate the man's accomplishments rather than swap gossip about him.
2 comments:
Re. the Turing test. If the machine hasn't passed the test until after some tweak, then why not say the tweak MADE it a thinker [assuming the idea that passing the test shows it is a thinker, which is problematic], rather than, as you have it, suppose one needs to infer [which would be ridiculous, considering how incompetent that pre-tweak machine might have been] that it was a thinker all along? Just saying...
Right... I always have to keep these posts too brief, but that's what I meant: it's ridiculous to think that a tweak could change the machine fundamentally from a non-thinker to a thinker. I was thinking of tweaks such as using contractions, adding slang, having a larger and more realistic set of canned responses, etc. You might be thinking of a tweak such as the magical whatever that allowed Skynet to achieve self-awareness in the Terminator movies... if that were a tweak, then I'd agree with you.
But on a side note, I think Turing was trying to get us away from thinking of AI in terms of consciousness. By defining machine intelligence his way, he sets the path for trying to develop it.
Post a Comment