Machine defeats man at go
A major milestone: a machine has consistently defeated very strong humans at at the game of “go” (report here ). The caveat is that this is on a 7×7 board, which is dozens of orders of magnitude less complex than the full 19×19 game.
Crazy Stone, the program in question, used the so-called Monte Carlo technique. Basically, it plays hundreds of thousands of random games and finds which move leads to the highest winning percentage. An interesting Wired article on this approach is here. A server allows bots to play thousands of games of each other, providing real-time information on the degree of promise of various approaches.
The real question is whether this is a one-trick pony useful only for tiny boards, or whether it can be usefully extended to 19×19. Until 20 years gives us immensely more powerful computers, we need some kind of abstraction to serve as the topic of Monte Carlo simulations. Or, perhaps Monte Carlo can be another trick in the go programmer’s bag of tricks, somehow combined with the opening books and connectivity analysis and pattern matching and heuristics that serve as the basis of today’s strongest 19×19 programs.
September 21st, 2006 at 09:54
please from where you have “a machine has consistently defeated a top professional at at the game of “goâ€?.”
thanks
September 21st, 2006 at 13:27
The headline is a bit misleading—9×9 is not a standard go board size. I don’t think this method will scale well to 19×19.
September 21st, 2006 at 15:00
Our AI lecturer once set us this problem as an assignment. He was so incompetent.
September 21st, 2006 at 15:20
Daniel, perhaps your lecturer was hoping that one of you was a genius. No such luck I assume.
http://www.snopes.com/college/homework/unsolvable.asp
September 21st, 2006 at 19:40
The real question is whether this is a one-trick pony useful only for 9×9, or whether it can be usefully extended to 19×19.
Yeah, that’s the real question, and the real answer is no, and has been for years. Nothing to see here, folks.
September 21st, 2006 at 23:05
Give it time, folks. To assume that a computer, or network of computers, can never defeat a human on a 19×19 go board is not the endorsement of human ingenuity you think it is.
September 22nd, 2006 at 05:29
By the way, if anybody wants to ponder what it means for us to be made obsolete by artificial intelligence, I recommend the book “Imaginary Magnitude” (“Wielkość urojona” in the original Polish) by Stanislaw Lem. Although the book was written back in 1973, and is presented as science fiction, it really made the subject come alive for me.
It’s a tough read until you see what the author is leading towards. Then it becomes unsettling.
Stanislaw Lem is usually remembered for “Solaris” and “The Futorological Congress”, but I think he will ultimately be best remembered for “Imaginary Magnitude”. Just don’t ask me who is going to be doing the remembering.
September 22nd, 2006 at 05:29
How much more difficult is 19 x 19 than 9 x 9? Even if the difficulty increases exponentially after the trival 3 x 3, it seems to me that the programmers and hardware designers are closing in on humans in this particular realm, and will soon surpass us.
Processors are becoming smaller and cheaper. Picture a parallel computer containing trillions of simple processors—a teraprocessor, you might call it. Now program in “imagination” and foresight in the form of internal simulations. Would a human mind, with it trillions of biological processors, be able to compete with that in any realm of the intellect?
So what is it that humans can do that cannot also be done by algorithms running in parallel? Can anybody here answer that without invoking something ineffable like a soul?
If your answer is something like “Love”, try actually defining what it is you are describing. You may find that you aren’t as sure as you originally were.
If and when computers learn to learn, we may find ourselves obsolete. We may also find ourselves wondering if we’ve been replaced by something alive or dead (if that actually means anything).
September 22nd, 2006 at 07:29
I could beat this program giving it 4 stones handicap on a 9×9 board. I cannot believe the story about the top go pro is true. The game follows.
(;FF[4]CA[Cp1252]AP[GoGui:0.9.1]SZ[9]KM[6.5]DT[2006-09-22]RE[W+Resign]
;B[ee];W[];B[ff];W[];B[dd];W[];B[cc];W[dg];B[fg];W[gd]
;B[eh];W[ce];B[ec];W[hf];B[ge];W[he];B[hd];W[gc];B[bd];W[dh]
;B[di];W[ci];B[hg];W[be];B[bi];W[ei];B[df];W[cf];B[gg];W[fb]
;B[bh];W[bg];B[id];W[hc];B[if];W[ic];B[ae];W[af];B[ig];W[ie]
;B[ch];W[eg];B[fh];W[di];B[fa];W[eb];B[gb];W[db];B[bb];W[fd]
;B[dc];W[ad];B[hb];W[fe];B[fc];W[ib];B[ef];W[cg];B[ac];W[cb]
;B[fi];W[ca];B[cd];W[ba];B[de];W[gf];B[ga];W[ea];B[aa];W[ha]
;B[ah];W[ag];B[gb];W[ai];B[bi];W[ab];B[bc];W[ae];B[bh];W[ah])
September 22nd, 2006 at 08:44
please from where you have “a machine has consistently defeated a top professional at at the game of “go�.� ?
thanks
October 28th, 2006 at 14:54
Part of the playing strength of Crazy Stone on 7×7 comes from an opening book that is based on an analysis done by humans. The 9×9 version does not have such an opening book and is much weaker than the 7×7 version.