mingwu wrote: > >> Playing randomly like that shouldn't work, but when you play Mogo et al, >> you see that intelligent behaviour emerges. >> >> > > Although interesting, I would hardly call that 'intelligence' :-) And I > suspect how far it can achieve on 19x19 board (compare to human players of > course). > > The explanations given were greatly simplified. Actually, these programs have a ton of research behind them are not just explained simply by a a bunch of random games. So the approach is quite intelligence.
One breakthrough has been making the play-outs "less random" and so they actually have been crafted to give the best results with a lot of outside knowledge added. Also, they are search based which means a search tree is built in memory and a great deal of effort has been spent on how to expand this tree in the most efficient way possible. This has already proved to be the best way forward (so far) as the 19x19 programs using the approach are playing the best go of all the programs. So you may worry about "how far it can achieve on 19x19", but now it's not the Monte Carlo approach that is really in question, it's the OTHER approaches that you should now be doubting. With the current trend of Moores law to produce faster computers with more cores and memory, this is truly good news for go programs and especially this technique which is guaranteed to improve with more powerful hardware. - Don _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
