Quest for AI: Go Talk

For those of you out there who follow my posts, you will know that I am very interested in artificial intelligence (AI). In fact, my question at Quest is “How should we create artificial general intelligence?”. (For those of you interested, artificial general intelligence is an AI that is good at many things, not just one thing). This year at Quest we have a club called Quest for AI. We share resources, work on projects, hold seminars, and give talks.

This past Thursday, some of us hosted an event for the Quest community to commemorate the historic victory of AlphaGo, an AI created by Google, over Lee Sedol, who at the time of the challenging was the top Go player in the world (he had recently become the second highest ranked Go player in the world, just before the contest with AlphaGo). This was an incredible achievement, reminiscent of DeepBlue’s victory in chess over Gary Kasparov in 1997. Chess is an incredibly complicated game, and has an incredibly large amount of possible game states. Go is even larger. For comparison, there are more possible game states in Go than there are atoms in the observable universe. Thus, AlphaGo’s 4-1 victory over Lee Sedol is quite amazing. Most experts thought that this level of play from an AI would be decades away.

A few members of Quest for AI, myself included, organized the talk. We got funding from the Student Representative Council for food, and hosted the event on the third floor of the library so that anyone who wanted to attend would be able to do so. The first speaker explained what Go was, how one plays it, and why it is such an interesting game. He then turned over the stage to me, and I explained the basis behind the machine learning technologies that AlphaGo used to learn Go, and how they all fit together to create an incredibly powerful player.

After the talk we all mingled, and played a game of Go. People were very curious about the technology underlying AlphaGo’s success, and it was great fun explaining how the different machine learning techniques work. Everyone seemed to have a good time. I hope to have more events like this in the future!

go1

Leave a Reply