Computers beating humans at chess is old news — hell, it happens to me every day — but the Chinese board game Go is infinitely more complex, which makes today’s news out of Google’s AI division DeepMind even more interesting.
DeepMind’s new Go-playing program, nicknamed AlphaGo, challenged the European Go champion to a series of matches in October and triumphed five games to nil.
There are more possible positions for Go pieces than there are atoms in the universe, which forced the team building AlphaGo to rely on watching how the world’s top players acted and then extrapolating from there.
According to this BBC report, AlphaGo’s runs simulated matches against itself millions of times to build the best results, learning from each match to find out what works and what doesn’t. AlphaGo will move onto Seoul in March to take on the world champion, Lee Sedol.
No comments on this article yet. Why not add your own?
You must be logged in to leave a comment.