Innovation

Google's AlphaGo to test the limits of human-AI collaboration at summit in China

Top Go players will meet in Wuzhen, China at the end of May for a Future of Go Summit, playing against, and collaborating with, Google DeepMind's AI platform.

screen-shot-2017-04-10-at-11-38-48-am.png

The world's no. 1 Go player, Ke Jie (far right).

Image: Google DeepMind blog


In one of the top achievements for AI, AlphaGo, Google DeepMind's machine learning platform, defeated Go world champ Lee Sedol in a game more complex than chess, in March 2016. AlphaGo beat Sedol in four of the five-game tournament, after mastering the game 10 years earlier than many experts predicted.

AlphaGo has continued to improve its skills throughout the past year, training by playing against top players from South Korea, China, and Japan. And at the end of May, AlphaGo will explore a new way to learn: By pairing up with top Go players and AI experts at the Future of Go Summit.

The five-day conference, held from May 23-27, will bring together AI experts, the China Go Association, AlphaGo, and China's top Go players in Wuzhen, China.

Here's a preview of the events:

  • Pair Go: Two Chinese pros will face off—with the help of an AlphaGo teammate, who will play alternating moves.
  • Team Go: A game between AlphaGo and a five-player team consisting of China's top pro players, working together to test AlphaGo's creativity and adaptability to their combined style.
  • Ke Jie vs. AlphaGo: The world Go expert, Ke Jie, will play against AlphaGo.

AlphaGo harnesses convolutional neural networks to play the ancient Chinese game. What has made it particularly impressive is the fact that it uses reinforcement learning, instead of being programmed specifically for the task. While IBM's Deep Blue achieved an AI victory in 1997 by beating world chess master Gary Kasparov, it was programmed with the moves. And Go is a complex game, with potentially 200 options per move, as opposed to 20 on a chessboard, relying heavily on intuition.

While it is interesting to see what AlphaGo is capable of, the summit in May will be a test for how AI and human collaborations can work. It will be a chance to see how human learning can be enhanced by AI. And the takeaways will likely extend past the gaming world. Manuela Veloso, head of machine learning at Carnegie Mellon University previously told TechRepublic that she was interested to see "if and how AlphaGo's learning approach may apply to other different 'non-game' problems." AlphaGo has already taken on the task of reducing energy use, and the technology has been applied to medical research projects as well.

SEE: Google DeepMind: The smart person's guide

Toby Walsh, AI professor at the University of New South Wales, also previously told TechRepublic that the nature of the game itself could change, and that the AI system "played moves that have surprised even Go masters."

"Will man and machine together be better than man or machine alone?" Walsh asked. "Each of us can bring unique strengths to the table."

The 3 big takeaways for TechRepublic readers

  1. At the end of May 2017, AlphaGo, AI experts, and top Go players will meet at the Future of Go Summit in China.
  2. The five-day conference will feature a variety of different scenarios that will involve human players and the AI platform playing with and against each other, with the goal of teaching AlphaGo to improve and to see what emerges from the collaboration.
  3. While AlphaGo is a particularly impressive AI achievement, the success of the machine learning platform applies to non-game arenas as well.

Also see...

About Hope Reese

Hope Reese is a Staff Writer for TechRepublic. She covers the intersection of technology and society, examining the people and ideas that transform how we live today.

Editor's Picks

Free Newsletters, In your Inbox