Google DeepMind AI finishes off Go champ in style
Google’s DeepMind AI team has taken the fifth and final round of its high-profile Go match with Lee Sedol.
The 18-time Go world champion had appeared to rally with a single victory in yesterday’s game, after the AlphaGo program had taken a decisive 3-0 lead.
However, the fifth and final game went the way of the AI once again for a 4-1 final score.
The last game was arguably the closest yet, however, with Lee Sedol managing to take it deep into overtime. DeepMind founder Demis Hassabis also admitted that the AlphaGo program made a “bad mistake” early in the match, but managed to claw its way back.
DeepMind’s success is seen as a major advancement in the field of AI. Up to now, no computer program has been able to beat a 9-dan Go player.
Related: Google’s AlphaGo and 5 more times machines destroyed humanity’s best
While Go’s rules are a lot simpler than Chess, the list of potential permutations on its 19 x 19 game board is vastly longer, making it impossible for a blunt force data-crunching approach to succeed. Proper strategic thought and a degree of on-the-fly intuition is required.
AlphaGo utilised deep neural networks and machine learning to gain an advantage that no computer program had managed thus far.
You can watch the final Go match here:
With the match complete, Google will donate its $1 million prize pot to charity.