[ExI] human level ai, was: RE: ai class at stanford

[spike]

Reiteration please, if the topic wanders off of Stanford AI class, do adjust
the subject line, thanks, since we gave that subject temporary papal
dispensation to sin, we want to retain righteousness in other areas.


>... On Behalf Of Kelly Anderson

>... If you want to prove that AI can't meet human levels of performance, I
would suggest starting by figuring out which class of impossibility it is,
and why.  -Kelly


The problem can be broken down into subcategories of human intelligence.  We
once thought chess required human judgment to reach the top levels, but
software blasted through the top human level and kept right on going.
Recently, the French Chess Association caught three of its own players, the
top players in that nation, cheating using computers.  Last week a major
tournament had a B rated player scalping masters.  They took away his
briefcase and pen, at which time he instantly reverted to B rated play, and
lost all the remaining games.  Two years ago, a cell phone won a major
tournament in Brazil which included two grandmasters, without even having to
call a friend.

One of the things I hope to get from the Stanford AI class is an idea for
why AI can or cannot be expected to meet and exceed human level performance
in the area of designing AI software.

spike