Artificial Intelligence Doesn’t Need to be Taught By Humans Anymore: AlphaGo Zero

Artificial Intelligence Doesn’t Need to be Taught By Humans Anymore: AlphaGo Zero

Martin Paev, CFA

Executive Director, SORTIS INVEST

Advances in Artificial Intelligence (AI) have reached a stage where computers don’t need to learn from human experience anymore.

DeepMind, the London-based AI company acquired by Google in 2014 has been working on AI software to master Go, the ancient Chinese strategic board game. Go is a two-player game invented more than 2,500 years ago, played on a 19×19 field, in which the aim is to surround more territory than the opponent. It is considered the oldest board game continuously played today.

AlphaGo, the program developed by DeepMind managed to beat Ke Jie, the Go world champion at the time, at the Future of Go Summit earlier in 2017 and was awarded professional 9-dan by the Chinese Weiqi Association.

After that, DeepMind continued the training of AlphaGo by letting it play millions of games against itself. A new version of AlphaGo was recently revealed, under the name AlphaGo Zero, which started learning Go from scratch without investigating any previous human games at all. In just 3 days it beat a previous version of AlphaGo by 100:0 games, in 21 days it exceeded the capabilities of AlphaGo Master which beat 60 of the top Go players in the world incl. Ke Jie. On the 40th day of its training AlphaGo Zero surpassed all other versions of AlphaGo to become probably the best Go player of all time.

The AI technology has huge potential to penetrate not just the gaming industry but all other areas of the human life.

The Earth in the XXII century might well be a world ruled by a virtual government powered by a cloud-based Artificial Intelligence software.

Leave a Reply

Your email address will not be published. Required fields are marked *