It’s 1-0 to the machines. The world’s best player of the ancient Chinese board game squared off against AlphaGo, the latest artificial intelligence-based opponent from my friends at AI company DeepMind.
Lee Sedol is pitting his wits against the machine in a match comprising 5 games in Seoul, South Korea, with a $1 million prize at stake.
He lost the first game today. I believe he will lose the match and, for the first time in history, we’ll be able to say that computers are better than people at what is generally considered to be the world’s most challenging board game.
Advertisement
Does AlphaGo’s ability signify that machines are on the cusp of human-like intelligence?
Intelligence is a slippery concept, but it is possible to break the requirements for an advanced, human-level AI into three main challenges: how it plans sequences of actions to achieve objectives; how it models its environment; and its ability to reflect on its environment and behaviour.
Three challenges
We are getting better at developing planning algorithms that are remarkably efficient and fast and it doesn’t look as if this requirement will be a bottleneck in the race to develop advanced AI.
We are also getting better at building systems capable of modelling challenging environments, but building a rich representation of the natural world, which such an AI would need, is a much bigger problem.
Watch the best human player take on the mighty AlphaGo AI at Go
Despite the complexity of Go and the subtlety of recognising promising positions of pieces on a 19×19 playing board with black and white counters, each of which can be placed in one of the 361 possible locations, it is still almost absurdly simple when compared with accurately representing the natural world. The difference between the two is a chasm of complexity that will take decades to cross.
However, it is the third requirement for human-like AI, reflection, that is the biggest sticking point. Reflection is the ability of a mind or a machine to weigh up its environment and its own behaviour and thought processes. The manifestation of this is what we often refer to as consciousness or self-awareness. We are still a long way from understanding why humans possess this extraordinary property.
And we are very much in the dark as to whether machines will one day be capable of it. It is, however, an intriguing area of exploration for both science and philosophy.
Mastering Go is a phenomenal achievement for artificial intelligence; another victory in an increasingly exciting area of research. However, when considered in light of the journey towards human level intelligence, it’s a relatively small step on a very long road.
In defeat, Sedol may find some solace in reflecting on that.
Read more: ‘I’m in shock!’ How an AI beat the world’s best human at Go
