In 1997, Kasparov was defeated by Deep Blue.
IBM was able to build such a program in part because they had the records of tens of thousands
of games that they could feed into to train Deep Blue how to play chess. You no doubt
know about AlphaGo defeating Lee Sedol at the game go. The platform on which AlphaGo
was built is called AlphaZero. It’s a much broader platform, designed to play games in
general, not go in particular, and it was recently tasked with playing chess. And in
just a matter of 24 hours, it developed to a level that could beat some of the best chess
programs in the world. But what’s interesting is that it is largely was not trained on historic
human games. And as such, as it learned to play on its own, it exhibited behavior that
would definitely be considered odd by chess standards. It would sacrifice pieces to gain
position on the board, it would say, place a queen way away in a corner, an unusual move.
The way it played chess was described actually as alien, that it played chess differently
than a computer program, and that it played it differently than a human; that it was an
alien form of chess. I wonder if there is a time when it will actually be regarded as
an alien. Artificial intelligence will not be seen not only as alive, but as a completely
different form of life, a completely different form of intelligence than human intelligence.
On the other hand, the program was made by people who figured out how to get it to do
what it does, and so the fact that it seems alien may just be an artifact of how it was
built. If you’re interested in artificial intelligence visit GigaOM.com or check out
my new book “The Fourth Age: Smart Robots, Conscious Computers, and the Future of Humanity”.