Microsoft Researchers Made An Xbox Game

For years, getting a computer to play chess has been a badge of honour for AI programmers. Getting an Xbox Live game to play something 10x more complex, then, seems like something to be proud of.

Microsoft has quietly released a first-party game on Xbox Live Arcade called "Path of Go", a computerised version of the ancient Chinese board game of Go. What makes this neat is that, as the video below points out, teaching a computer to play chess is relatively easy. You can only make 20-30 moves at any one time. But in Go, there can be up to 361 moves possible at any one time, making a realistic simulation of the game difficult.

A team from Microsoft's research labs has managed it, though, using the new "F Sharp" programming language (and the results of 250,000 actual matches) to create a credible, believable AI opponent for the game. While there are plenty of other Go simulators out there on the market, the strict nature of their programming makes them suitable only for experienced players; Microsoft's game, meanwhile, is pitched at newcomers, claiming to play more like your average human than a dedicated machine.

It's rare that a Microsoft research project is turned into a retail product, but because this was so simple and practical, it's now available for sale. Microsoft hopes that the AI programming used in the game can now be applied to other, more complex projects, including video games.

Amazingly, Path of Go has been in development since 2004, yet was only released in December 2010. It uses Xbox Live Avatars, and costs USD$5 (400 MSP).

[via Slashdot]


    Those research guys really do some amazing things... it's a shame so little of it gets beyond prototyping.

    See I would probably be down with that on my iphone or maybe PC, but if I am going to game on my big plasma and surround sound I will always opt to play something more awesome tan a board game.

    A normal Go board is 19 x 19, whereas the board in that screenshot is only 9 x 9. I had to make an AI play go on a 9 x 9 board for a 3rd year AI assignment at university. The possible move set for a 9 x 9 board is actually fairly manageable. A serious Go AI is WAY more complicated than what this game would have. Still, if it can play well on that size board it's doing better than my AI did. I'm not claiming I could do better, but I was actually kind of hoping that they'd made something amazing until I noticed how small the board was.

Join the discussion!

Trending Stories Right Now