Something strange happened in the world of artificial intelligence (AI) on Wednesday.
Hours later, DeepMind — a startup based in London that was bought by Google for £400 million in 2014 — said it had already developed an AI named AlphaGo that had just beaten the best Go player in Europe.
DeepMind’s breakthrough was splashed across the front cover of science journal Nature yesterday evening and covered by over 200 media titles.
“This is the first time that a computer Go program has defeated a human professional player, without handicap, in the full game of Go – a feat that was previously believed to be at least a decade away,” explained the DeepMind research paper — Mastering the game of Go with deep neural networks and tree search.
At first, the DeepMind announcement appeared to make a mockery of Zuckerberg’s brag earlier that day.
Business Insider understands that Facebook may have gotten wind of DeepMind’s announcement and rushed out an update to an existing AI research paper, complete with several typos. But Facebook’s AI agents can only beat amateur Go players, not professionals.
In December, Facebook AI researcher Yuandong Tian told Wired there was a “friendly rivalry” between Facebook and DeepMind, saying there would be “pride” for the company who conquered Go first.
A Business Insider source said a number of people in Facebook’s AI team were aware that DeepMind’s announcement was coming.
“There were rumours that this was happening, and they (Facebook) were aware of those rumours as well so I believe he (Zuckerberg) knew,” they said.
A journalist on Twitter also said they’d been tipped off about DeepMind’s announcement ahead of Zuckerberg’s Facebook post.
Either way, it’s a major breakthrough in the field of AI for a British startup that’s been operating quietly from a base in King’s Cross, London.
Go is one of the most complicated games in the world, with trillions of possible moves. The fact that a machine can now beat the world’s best humans at the centuries-old Chinese board game could be a significant milestone in AI.
Eventually, the DeepMind team hopes to use its AI to solve problems in the real world, whether through making medical diagnoses or by modelling the climate.