AlphaGo - the AI ​​that once defeated the world's No. 1 Go player was officially usurped

Once defeated Lee Sedol - the world No. 1 go player - in 2016, AlphaGo is now defeated by AlphaGo Zero, a new AI created by his father Julian Schrittwieser.

A few years ago, when Julian Schrittwieser joined the Google-owned artificial intelligence company DeepMind, the Go game was often called the "Holy Grail" of machine learning. This two-person antagonistic game originated from ancient China is virtually unrestricted by any rule and if you want to win, the player's intuition is an extremely important factor. At that time, it was anticipated that it would take a decade before the AI ​​could defeat the world's leading Go players.

But in March 2016, a program developed by Schrittwieser and his associates at DeepMind was brilliantly defeated, the defending champion of the world at that time - in a prolonged match 5 Inning, with 4 - 1 results, attracted more than 100 million followers. Go fans call this "the game of the century".

That's not enough, Schrittwieser and colleagues continue this success with a new achievement much more impressive. In October 2017, their new program, AlphaGo Zero , destroyed the seniors program AlphaGo with a shocking ratio of 100 - 0.

Picture 1 of AlphaGo - the AI ​​that once defeated the world's No. 1 Go player was officially usurped
AlphaGo Zero destroyed the elder program, AlphaGo, with a shocking ratio of 100 - 0.

Unlike AlphaGo, which learned to play Go by studying human matches, AlphaGo Zero learned by playing with itself - a special ability imprinted with AI.

"With AlphaGo Zero, we find that even in areas where we do not have human knowledge, we can still create that knowledge and let the system learn from itself," Schrittwieser said.

Schrittwieser is an Austrian citizen, currently leading the software engineering team in the AlphaGo Zero project. He is also the leader behind the 3rd DeepMind initiative - AlphaZero, a more universal algorithm, with transcendent Go, chess and Shogi (Japanese boardgame) skills. Promoting universalization, according to Schrittwieser, is the key to DeepMind's mission to build intelligent machines independent of intuition, thereby providing better solutions to problems. but the approach is often affected by human prejudice. Schrittwieser believes that this could lead to completely new improvements, implemented by AI, in areas ranging from pharmaceuticals to materials science.