I’m currently trying to move from Southampton to London, and it’s a pain having to trawl through the piles of stuff accumulated over several years. Many things have been thrown out, and should have been thrown out years ago, but some things are rediscovered on the process of moving. Like the game of Go (invented more than 2500 years ago in China), and the rather too posh set I bought myself only a couple of years ago. I decided that I wanted to play, to learn to play, properly, after having dabbled with it too many years earlier.
I gave the game a certain cachet when, as an undergraduate, I first discovered it, being played by two of my lecturers, John Washbrook and Simon Peyton Jones, at lunchtime. Only later did I begin to understand the inherent difficulty of Go, as opposed to other games like Chess, especially for machines.
And this week, coincidentally, I read in The Economist (27 January 2007) that the performance of machines is now improving, by using Monte Carlo techniques. For example, the MoGo system is apparently ranked 2,323rd in the world, and in Europe’s top 300.
Not sure it’s worthwhile learning to play now!
Well Mike, perhaps you can learn to play it just so you could develop a better agent player for it. Considering the inherent complexity of the game, traditional brute-force-based search methods like the ones used in chess will not go very far, and I do think that the level of randomness of Monte Carlo methods may also limit what can be achieved on that front.
Comment by Felipe Meneguzzi — February 1, 2007 @ 10:18 am
Ask Steve. He knows how to play it: he did his masters on it
Comment by Ana — February 12, 2007 @ 10:29 pm