Monday, December 26, 2005

Mis-nagid's Take on Turing

Whenever Yaakov Menken attempts to write about science hilarity follows. Though the chuckles most often come from his impassioned defense of Inteligent Design, he fares no better when he attempts to explain Artificial Intelligance.

The Turing Test and the Limits of Science by Yaakov Menken

For more on the story, we go to DovBear's science correspondent: Mis-nagid:

Menken can't even get the title right. Failure to pass a Turing test has little to do with the limits of science. (Menken has now shown twice that he has no idea what science is.) The failure says more about the overoptimism early AI pioneers had before they realized how hard the task was than about anything else. For example, Marvin Minsky, a founding father of AI, once gave his star pupil Gerry Sussman the task of writing a program that can describe the contents of a photograph -- as his summer project, in 1966. The lessons from the seeming intractability of full-blown AI are all about engineering hubris, not the limits of science. Not that there hasn't been advances in AI; just ask Gary Kasparov. Or buy a Budweiser, QA-ed at high speed by computer vision.

Nor does Menken really understand Alan Turing's work. Turing's original paper in which his test appears makes it clear that the test wasn't about the development of consciousness or the ability to think. In fact, the test was only mentioned in passing as an interesting thought experiment about how to detect intelligent behavior. This is made clear by the fact that in the original paper (where it was called the imitation game) the example is not to detect a computer, but to detect which of the two people you are teletyping to is the man and which is the woman! The paper then continues:

"We now ask the question, 'What will happen when a machine takes the part of A in this game?' Will the interrogator decide wrongly as often when the game is played like this as he does when the game is played between a man and a woman? These questions replace our original, 'Can machines think?'"

Turing's test was replacing the question of computers thinking, not asking it. Ironically, Menken's
own link mentions that in its very first sentence: "[...]Alan Turing, in an attempt to develop a working definition of intelligence free of the difficulties and philosophical pitfalls of defining exactly what constitutes the mental process of intelligent reasoning, devised a test, instead, of intelligent behavior." So why's Menken complaining that "not one [program] is an actual attempt to get a computer to think for itself?" Because he doesn't understand the Turing test.

Menken digs his hole deeper by comparing a computer scientist's thought experiment with a rabbinical legend. Yeah, a Turing test is exactly like a mythical creature, except where they're totally different. It's even funnier because he seems to believe the magical story as history. In that one sentence he shows his ignorance of both biology (golems?) and computer science, with the guts to finish it off with the flourish that he is "less sanguine" about the prospects of those exact topics. Menken should worry less about artificial intelligence developments and work on developing some of the natural kind.


P.S. Turing was a gay man who saved Britain in WWII with his work in cryptanalysis. His work was so secret that he never told anyone that he was a war hero even as his own country was persecuting him in its courts for being gay. They forced him to get injections of hormones to "cure" him, and he ended up commiting suicide. Throughout it all he never told anyone what he had done, not even to save himself from the country he had served so honorably.

I wonder what Menken thinks of that.

Sometimes, I wonder if Menken thinks at all. Thank you Mis-nagid, for that report. More in a moment.