AlphaGo, Google’s AI becomes the best Go player in the world by winning three games against world number one, Ke jie. AlphaGo once battled other champions like Fan Hui and Lee Sedol, which allowed him to improve, in addition to millions of games played against himself.
Twenty years ago, Deep Blue, IBM’s supercomputer, defeated world champion Garry Kasparov with his algorithms and great computing power, sweeping all the gameplay on many shots.
But faced with the game of Go, which has an immense number of possible combinations, the computing power is not enough, it is necessary to improve the algorithms.
Two methods are used in AlphaGo: the Monte Carlo method and Deep Learning. The Monte Carlo method makes it possible not to scan all the configurations of a tree of blows and the Deep Learning allows AlphaGo to learn on its own. Feeding AlphaGo with so many parts and playing it against itself has helped greatly improve AlphaGo in a minimum amount of time.
For AlphaGo, the latter consisted in feeding the program of many games and making it play against itself. This allowed him to make considerable progress in an extremely short time.
After emerging as the new world champion in the game of Go, AlphaGo will retire and AlphaGo developers will now work on general algorithms for medical and scientific research to combat certain diseases, reduce the consumption of energy or invent new materials.
The AI is an asset in this kind of project because it makes it possible to process terabytes of heterogeneous data at high speed and to extract its meaning.
AlphaGo’s advanced algorithms will therefore make it possible to advance many areas such as the medical sector by fighting certain diseases or the energy sector by reducing energy consumption.