More and more, we’re hearing stories about artificial intelligences that are “learning to learn”—that is, adopting a learning style more elastic than an algorithm that might possess more data but less creativity. Facebook is adding to that conversation with a recent blog post from Mark Zuckerberg about how they’re developing an AI that can play the 2,500-year-old Chinese game of Go. The thing is, Google got there first.
The ancient pastime of Go is one of the rare games in which humans still consistently beat AIs, due to the fact that there are estimated to be about 10 to the 761st power moves, more than the number of atoms in the universe. In a recent Facebook post, Zuckerberg explained the Facebook AI Research team’s goals and progress:
Scientists have been trying to teach computers to win at Go for 20 years. We’re getting close, and in the past six months we’ve built an AI that can make moves in as fast as 0.1 seconds and still be as good as previous systems that took years to build.
Our AI combines a search-based approach that models every possible move as the game progresses along with a pattern matching system built by our computer vision team.
That same day, Google announced the creation of AlphaGo, an AI that can beat human players at Go:
Traditional AI methods—which construct a search tree over all possible positions—don’t have a chance in Go. So when we set out to crack Go, we took a different approach. We built a system, AlphaGo, that combines an advanced tree search with deep neural networks. These neural networks take a description of the Go board as an input and process it through 12 different network layers containing millions of neuron-like connections. One neural network, the “policy network,” selects the next move to play. The other neural network, the “value network,” predicts the winner of the game.
We trained the neural networks on 30 million moves from games played by human experts, until it could predict the human move 57 percent of the time (the previous record before AlphaGo was 44 percent). But our goal is to beat the best human players, not just mimic them. To do this, AlphaGo learned to discover new strategies for itself, by playing thousands of games between its neural networks, and adjusting the connections using a trial-and-error process known as reinforcement learning. Of course, all of this requires a huge amount of computing power, so we made extensive use of Google Cloud Platform.
First checkers, then chess, now Go… but most interesting is that both companies had the same idea, only for Google to one-up Facebook. I’ll be curious to see if Facebook’s AI Research team continues forward on their AI—and if the Facebook AI and AlphaGo will ever be pitted against one another.