Don's reply is inspiring. So I would make it public to the list. :)
Aja
On Wed, Jul 13, 2011 at 9:11 AM, Aja <[email protected]> wrote:
Thanks Terry's clarification. Indeed, I misunderstood what Olivier meant.
Hope it didn't mislead any one.
By the way, Don, you know about chess more than us. I wonder if chess is
close to be solved now? Already 14 years since Deep Blue defeated Kasparov....
Chess is a very long way from being solved and we are not even approaching
perfect play. If I had to put a figure on it I would put a lower bound of 800
ELO away, but I think it's much more than this.
I don't think most people realize that in chess humans are not very strong and
neither are computers. It is now the case the computers are stronger than the
best humans by perhaps 200 ELO. Every year we see programs come out that are
50 ELO or more stronger than the previous best version, so we do not appear to
be very close to the top.
Being a top chess program author I see too many problems even with my own
program Komodo. If I can critique the evaluation function and see problems
then we cannot be very close to the ultimate player. To put this into
perspective Komodo (my program) is in a pack of programs that are within 70
ELO of each other at the top of the rating lists, which means any of these
programs can beat any other at least 40% of the time. The top program in
this pack is Houdini and it is in clear first place at the moment, the rest are
bunched together.
Since I know that my program sucks, I can say that we have hundreds of ELO to
go!
Don
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
