A world Go champion’s reflections on AlphaGo

Editor’s word: Lee Sae Dol is an expert Go participant who performed a now-famous public match towards AlphaGo, Google DeepMind’s AI system, again in 2016.

As a world Go champion, I’d solely ever performed the sport towards different people. That modified in 2016 once I performed a five-match recreation of Go towards AlphaGo, Google’s AI system. I admit, I underestimated simply how highly effective AI might be and ended up solely successful one out of our 5 video games collectively.

Once I first received the invitation from Google to play against AlphaGo, I didn’t fairly grasp how severe of a matter it was. I believed it might be a straightforward victory, and that this was an informal experiment. However as soon as this recreation grew to become public, I spotted how large of a deal it was.

And I nonetheless keep in mind how amazed I used to be at how brilliantly AlphaGo performed, the strikes it got here up with. Go is a deeply complicated strategic recreation — famously way more difficult than chess, with 1,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000 doable board configurations. Meaning AlphaGo must be inventive, not simply calculative.

Later, I realized that scientists had predicted AI wouldn’t attain that sort of functionality for one more decade. And this 12 months marks the eight-year anniversary of the sport I performed towards AlphaGo. In these eight years, AI has developed at an unimaginable velocity. Go gamers world wide have now used AlphaGo to attempt to uncover new strikes, new methods and new concepts on this historical recreation.