Human+AI vs AI+AI

G

Guest

Guest
Archived from groups: rec.games.chess.misc,rec.games.chess.computer (More info?)

I wonder if someone has done a reasonably scientific experiment
by letting a decent human chess player use a strong chess program
in a match against another copy of this very same chess program.

In particular, I'd like to know if, say, a 2400 GM using a 2400 AI
is able to gain strong advantage against the AI. To make the match
more fair, the resources for the opponent AI could be doubled, so
as to simulate the AI being able to use another copy of itself.

I think such experiments could tell us a lot about chess and AI:
if the GM+AI starts winning easily against AI+AI, then this could
indicate that human insight into chess is still relatively very
deep; on the other hand, if the GM does not upset the rating in
such a situation, then perhaps human chess insight does not make
much difference anymore.

- D
 
G

Guest

Guest
Archived from groups: rec.games.chess.misc,rec.games.chess.computer (More info?)

It's all too known that in the form of advanced chess (human+AI)
any experienced club player has easy time against both top human
and top computer players. I can beat any engine on my computer
with the help of much lower engine, for instance I have tried
me+cm3k and Fritz 8 can give no resistance at all against
us - and I'm just 1900+ rated. Although, there exists no engine that
I haven't beaten barely by learning method, playing the same openings
that I know very well again and again and finding the path to
victory. For instance cm10th ed, fritz 8, junior 8 have all
different type of vulnerability to my stonewall attack. With black
pieces it's not however all that clear, for now I have achieved best
results for now with different f5-variations, i.e. dutch defence
when playing black.