[lbo-talk] Kramnik vs Deep Fritz

ravi ravi at platosbeard.org
Tue Nov 28 10:16:51 PST 2006


At around 28/11/06 1:31 am, joanna wrote:
> If it makes you feel any better, it's not important that the computer
> can win.
>

No worries -- I was not feeling bad about the computer winning. I was just taken aback by the type of error Kramnik committed.


> After all, no computer now or ever will be able to create the game of
> chess.

I am afraid I believe in a mostly mechanistic version of human thought: I think it is not impossible that a computer may invent a game such as chess.

Only a bit related:

There is not a lot of evidence of that sort of capability in computers today. Back in the 90s (IIRC) a fellow named Doug Lenat wrote an AI type program called "AM" (based on "I think therefore I am"), which was armed with a few atomic mathematical concepts and axioms, along with some inference rules, and heuristics (for guessing on when to proceed and when to give up, etc). Lenat let the program run infinitely, spitting out what it considered interesting results. To the mild surprise of many it did discover (not invent, unlike chess) many interesting mathematical laws and theorems (such as the unique factorisation theorem), but soon got bogged down in fruitless byways and circular paths. At one point however, it started exploring highly composite numbers (the conceptual opposite of a prime number. Example: 24) which are usually not of much interest to many (including mathematicians). In doing so, it echoed the behaviour of a rather eccentric mathematician, Ramanujan, and again surprisingly, rediscovered many of his results.


> Humans may be flawed, but as Rumi put it, the light can only come in
> through the cracks.

That's a great way to put it.

--ravi



More information about the lbo-talk mailing list