The story of AlphaGo

In 2016, the computer program AlphaGo captured the world’s attention when it defeated the legendary Go player Lee Sedol. The ancient board game of Go is one of the most complex games ever devised, with more possible board configurations than atoms in the universe. It was a longstanding grand challenge for artificial intelligence and AlphaGo’s 4-1 win was considered by many to be a decade ahead of its time. The system was invented by DeepMind, co-founded by scientist Demis Hassabis. Five months earlier, AlphaGo had beaten European champion Fan Hui, becoming the first program to defeat a professional player.

Over 200 million people around the world watched as a legendary Go master took on an unproven AI challenger for the first time in history.

During the matches AlphaGo played a series of moves which challenged conventional thinking about the game.

And when the black stone touched the Go board at move 37 of game two, something magical happened. ‘I think we’re seeing an original move,’ one of the match commentators said, double-checking his screens to see if there had been a mistake.

“Move 37, played by AlphaGo during game two of our five-match tournament against the genius Lee Sedol – widely considered to be the greatest Go player of the past decade – turned out to be the critical move that helped AlphaGo win the game. But for me and many of the people watching, it was not simply the fact that AlphaGo won, but how it won that we found most fascinating.

In Asia, Go is considered an artform, not just a game. But it’s ‘objective’ art. Anyone can play an original move on a Go board by simply playing randomly. Yet a move can only be considered truly creative if it’s also effective. In that sense, Move 37’s decisive role in game two represents a moment of exquisite algorithmic ingenuity that not only changed the game of Go forever, but also came to symbolise the enormous creative potential of AI.”

Demis Hassabis, DeepMind co-founder and CEO

"Our ambition at DeepMind is to build intelligent systems that can learn to solve any complex problem without being taught how - and then use these systems to help find solutions to some of society’s biggest challenges and unanswered questions. Put another way, we want to solve intelligence and then use it to solve everything else.

Perhaps surprisingly, we use games as a way to evaluate our systems, and the most complex and subtle board game that exists is Go. It is a game of intuition rather than calculation. Top players often say they played a move because it ‘felt right’ rather than because it was part of an explicit plan.

So, when we began our research, we knew we would need a novel approach. Traditional games engines comprise thousands of rules handcrafted by strong human players that try to account for every eventuality in a game. The final version of AlphaGo does not use any rules. Instead it learns the game from scratch by playing against different versions of itself thousands of times, incrementally learning through a process of trial and error, known as reinforcement learning. This means it is free to learn the game for itself, unconstrained by orthodox thinking."

Demis Hassabis, DeepMind co-founder and CEO

"Our first opportunity to test AlphaGo was against European Go champion Fan Hui in October 2015. While Fan Hui played with immense skill, AlphaGo won in five straight games, marking a world first for a computer Go system playing against a human professional.”

Demis Hassabis, DeepMind co-founder and CEO

“I was very confident when I first arrived at the DeepMind offices in 2015 to play AlphaGo – after all, no system had ever come close to defeating a professional Go player and it was unthinkable to me that that would change. Go is not only complex but requires creativity and intuition – traits not commonly associated with a computer program. So when I lost 5-0, I found it very difficult and it made me worry about the future of the game. But now I see that my match wasn’t the end; it was the dawning of a new era for Go and one that I am thankful to have experienced first-hand.”

Fan Hui, professional Go player

“Move 37 goes against all conventional teaching and no experienced human player would ever have played it. In fact, we know from AlphaGo’s calculations there was just a one in ten thousand chance of a player selecting that move. It was a moment of inspiration that came from its unique approach to the game. Unlike the way I – and all other human players – approach Go, its decisions are unencumbered by the tradition, theory, and teaching of human play. Instead, it learns the game for itself, giving it the opportunity to for fresh thinking and leading to a unique ‘free spirited’ style which in turn has unshackled human players from tradition and allowed us to also think differently about the game.”

Fan Hui, professional Go player


“Move 37 might be the first glimpse of a bright and bold future where humanity harnesses AI as a powerful new tool, helping us discover new knowledge that can solve some of our most pressing scientific problems.”

Demis Hassabis, DeepMind co-founder and CEO