That's a good way of putting it. AFAIK, Mogo has nothing comparable to the long-term memory of human players; each time it approaches a given joseki, it must recompute the right play from scratch.
Aeons ago, I read a few papers by cognitive scientists at CMU - Newell and Simon, if memory serves - who looked for ways to "chunk" bits of knowledge in order to speed processing. Their theory was that, with practice, humans build chunks which encapsulate their experience into more efficient models. Hats off to those who posited that the problem really is scalable. I admit to having been sceptical -- I believed there were some inherent problems which would limit the degree of scalability. Evidently, given enough time and processor power, programs like Mogo really can and do achieve dan-level play on a 19x19 board. <eagerly awating the day when my desktop computer has that power> Terry McIntyre <[EMAIL PROTECTED]> “Wherever is found what is called a paternal government, there is found state education. It has been discovered that the best way to insure implicit obedience is to commence tyranny in the nursery.” Benjamin Disraeli, Speech in the House of Commons [June 15, 1874] ----- Original Message ---- From: Peter Drake <[EMAIL PROTECTED]> To: computer-go <computer-go@computer-go.org> Sent: Friday, August 8, 2008 1:48:23 AM Subject: Re: [computer-go] Re: mogo beats pro! Yes, MoGo gained much more from the longer time setting than Mr. Kim did. Note that Mr. Kim used very little of his time in the one-hour game. He said after the match that using more time would not have helped him. This is an interesting property of Monte Carlo Go. At the risk of overgeneralizing, it may be that digital computers have an advantage over brains in terms of fast and accurate short-term memory (dare we call it "concentration"?). A human pro has better lightning instincts (fuzzy long-term memory?), but some time for MC sampling allows the program to develop that instinct (or something analogous) over the course of the game. If we could only decide what to store (and how to store it) between games, we could get good blitz performance and superhuman long game performance. Peter Drake http://www.lclark.edu/~drake/ On Aug 7, 2008, at 10:26 PM, Dave Dyer wrote: My take-away from watching the match is that blitz performance wasn't at all representative. A human playing blitz games might do 90% as well as at a full length game, whereas mogo's performance looked like it scaled more linearly.
_______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
