[FOM] AI in Chess/Kasparov
Timothy Y. Chow
tchow at alum.mit.edu
Mon Feb 1 11:43:47 EST 2010
Harvey Friedman <friedman at math.ohio-state.edu> wrote:
> Most interesting is just how much better human + computer is than
> human alone or computer alone.
In the case of chess, this is starting to become a thing of the past also.
Advanced Chess has not really taken off because the amount that the human
can contribute keeps shrinking. The freestyle chess event mentioned by
Kasparov took place in 2005, two years before Rybka took the World
Computer Chess Championship by storm. Rybka is the current top dog, and
while it's still obviously not perfect, based on the current trend, the
smart money is that future improvements in playing strength will mainly
come from the silicon side and not from carbon-silicon hybrids.
A few years ago I made a post to FOM on a related topic.
If one is interested in carbon-silicon collaborations, I think go is a
much more promising game than chess. Until recently, go programs were
hopelessly bad. In the past couple of years, however, the use of
so-called "Monte Carlo methods" where the computer evaluates a position by
playing it out to the end of the game randomly (!) has significantly
improved the programs' playing strength.
I believe that state-of-the-art machines can now beat top professionals
with only a seven-stone handicap. I expect that it will not be too long
before the carbon-silicon partnership that Harvey is intrigued by will
become a reality in the go world.
More information about the FOM