The Evolution of Friendly Competition Between AI and Humans

Google's AlphaGo recently defeated the legendary 18-time Go World Champion, Lee Se-dol in the first two games of their series. We decided to go back to the origins of AI versus human gaming.

IBM's AI Champions

Pushing the limits of friendly competition between humans and AI probably fell into the realm of popular culture in 1997, when IBM's Deep Blue defeated Garry Kasparov, the World Chess Champion at the time. There is some controversy surrounding the event, which was covered in the film Game Over: Kasparov and the Machine, but regardless of opinions on the situation, an AI playing chess at such a high level is a great accomplishment.
         

Deep Blue versus Garry Kasparov (Image from NBC News)

Deep Blue could evaluate 200 million positions per second, which was especially impressive for 1997, but the AI was limited to the parameters of the game of chess. IBM later took this idea to the next level with Watson, which came into the media spotlight when it won Jeopardy! against two former champions. The idea behind Watson was to create a computer that can teach itself new things, in order to use it's massive computing power to contribute process information like a human does. Since Jeopardy! encompasses a wide array of subject matter, and the questions asked on the show haven't been asked yet, Watson could not just find the answers by scanning through text. It would have to think like a human, by putting together clues from context. We've been hearing about Watson ever since.

IBM's Watson, the new overlord of Jeopardy!

 

A New Player Enters

AlphaGo, an AI program developed by Google's DeepMind recently accomplished another feat in the realm of gaming. The AI program defeated the 18 world champion Lee Se-dol in two games of Go, an ancient Chinese board game that is more complex than chess due to an infinite amount of possibilities for moves. Since there is no finite set of data for moves, AlphaGo has to make decisions based on intuition instead of drawing from a database the move that is most strategic based on probability. The five-game series isn't over yet, and Lee still has a chance to come back and win it all, but the odds aren't in his favor.

Lee was confident in his abilities, before the matches, but eventually had to tip his hat to AlphaGo. He had this to say about the games in a post-game press conference:

"Yesterday I was surprised but today it's more than that — I am speechless, I admit that it was a very clear loss on my part. From the very beginning of the game I did not feel like there was a point that I was leading."

Win or lose, Lee deserves credit for accepting the challenge and competing with AlphaGo, especially considering how human versus AI gaming competitions have gone in the past. Stay tuned for game 3, which begins tonight at 10:30PM US Eastern time. You can find a live stream of the upcoming game and watch the first two games on DeepMind's Youtube channel.