Science & Technology

Google invents an AI that learns on its own

By T.K. Randall
October 18, 2017

Image: AI-generated (Midjourney) The latest creation of Google's artificial intelligence group DeepMind, the new software has not only managed to learn how to play the ancient Chinese board game 'Go' all on its own, but has even developed its own new moves and strategies - all within the space of just three days.

When pitted against its predeccessor AlphaGo, which famously beat South Korean grandmaster Lee Sedol back in 2015, the new and improved AlphaGo Zero managed to win the game 100 points to 0.

Unlike AlphaGo which learned by analyzing previous matches between human opponents, AlphaGo Zero mastered the game by simulating millions of games against itself to develop its strategies.The team behind the system is now using it to help solve real-world scientific problems.

"For us, AlphaGo wasn't just about winning the game of Go," said DeepMind CEO Demis Hassabis.

"It was also a big step for us towards building these general-purpose algorithms."


Source: The Guardian

Tags: Google, Artificial Intelligence