Anna Rudolf Visits The AlphaZero Headquarters

So the day came. On December 11 I was allowed to sneak into the place where magic happens: the DeepMind headquarters in London, the home of the chess-playing project AlphaZero.

AlphaZero is the general purpose artificial intelligence system DeepMind built and it blew my mind from the moment it mastered chess as part of DeepMind's experiment with the algorithm back in 2017. The sheer thought that a machine managed to learn chess with no input but the rules, simply by playing games against itself and taking conclusions from them to gradually increase its understanding and playing strength to heights way above human level...is mindblowing. And it took AlphaZero four hours in total.

An academic paper on this tremendous chess (and shogi and Go) player has recently been published in the Science journal, shedding more light on the way AlphaZero thinks, learns and plays. DeepMind also released new, never-before-seen games from the matches the AI system fought against the strongest chess engine, Stockfish. All this on December 6. Two days later I packed my bag and flew to London.

Easy for us, chess people, to just pop into a chess event, namely the London Chess Classic, and run around with a camera. Not so easy at all to be allowed to take the elevator one floor below. (The semifinals of the London Chess Classic were held on the eighth floor of the Google HQ at King's Cross, the offices of DeepMind are on floors six and seven of the same building).

My gratitude goes to the DeepMind team for giving me special permission to invade and film in this exclusive space, and shout-out to the 15+1 participants who took part of my unofficial survey. Here's the visit, here's the challenge: