In March 2016, the computer program AlphaGo cracked the game of Go, an ancient strategy board game with ~2.082 ⨉ 10170 legal positions (imagine a 2 followed by 170 zeroes). The complexity of Go has made it the subject of Artificial Intelligence (AI) research for decades, but until last year computer Go programs were no match for human intuition.
This talk will give a gentle introduction to the machine learning techniques that made the breakthrough possible, along with links and resources to open source machine learning libraries.
AlphaGo is a computer program developed by Google DeepMind. Its algorithm uses a Monte Carlo tree search to find its moves based on knowledge previously «learned» by machine learning; specifically by an artificial neural network given extensive training, both from human and computer play.
Google DeepMind, a London-basedartificial intelligence company founded in 2010, and acquired by Google in 2014. They famously developed a neural network that learns how to play video games in a fashion similar to that of humans (see «Google's DeepMind Masters Atari Games», Forbes, Feb. 2015).
DeepMind has released their flagship product, DeepMind Lab, as open source software. Based on based on id Software's Quake III Arena, it is a fully 3D game-like platform tailored for agent-based AI research, and observed from a first-person viewpoint, through the eyes of the simulated agent.
About the speaker:
Eric Rasmussen is an avid Go player, functional programming evangelist, and Lead Software Developer at newcars.com.