Cracking Go

prostoalex writes "IEEE Spectrum looks at current trends in artificial technology to crack the ancient Chinese game of Go, which theoretically has 10^60 potential endings. Is conquering the game via exhaustive search of all possibilities possible? 'My gut feeling is that with some optimization a machine that can search a trillion positions per second would be enough to play Go at the very highest level. It would then be cheaper to build the machine out of FPGAs (field-programmable gate arrays) instead of the much more expensive and highly unwieldy full-custom chips. That way, university students could easily take on the challenge.'"

Comments

Because humans and computers are different

[Sycraft-fu]

The way humans deal with data is just different from how computers do. Tasks we find trivial they find hard and vice versa. For example if you show a human pictures of cats and dogs and ask them to classify them according to which animal they are, that's easy for us. For a computer, it's a very hard task. However a computer has no problems dealing with an analyzing a 50,000 item long table of numbers, whereas a human needs them converted to another form to make sense of them.

Well, when it comes to solving games, it's a similar deal. Computers aren't good in the way humans are. They are good by mapping out possibilities, tons of them, and then figuring out based on that. That's how good chess programs work. At their core it is more or less plotting all the possible endpoints several levels deep from a given move. They then calculate the most favourable set and that's how they choose their move. A little more complex than that, of course, but at its core it is simply mapping out billions of board positions and using those to decide moves.

Thus a similar thing probably would work well for Go, especially if you can do an exhaustive search and go to literally ALL endpoints for every move. It also has unique strengths. Many good human players work off of psychological aspects of the game. Kasparov, for example, was known for "throwing lightning bolts" at the game board, doing things other people couldn't cope with. That doesn't work for a system that just analyzes the current position and where it can go from here. There's no way to confuse something like that, no way to play to a mental state. It simply weighs what you've done, and what you could do from here.