Go, Going, Gone: AI just beat the world champion in the world

Arya Sharma/Catch News

Go, Going, Gone: AI just beat the world champion in the world's toughest game

Let's observe a moment of silence for Lee Se-dol. Lee is a man who dedicated his life to something few people even care about, the ancient Chinese board game 'Go'. Being South Korean he doubtless fought against the odds to pursue his passion. Probably even as his parents looked on in massive disappointment as his next door neighbour Cho joined a mass-brand tech-company.

Still Lee soldiered on, accompanied I imagine by an uplifting musical score and a harsh but lovable mentor until finally, he was the world's best... Until today, when he was beaten by Google's artificial intelligence (AI).

Also read: Microsoft's new software can tell how you're feeling from a photo

While Lee has only lost two rounds in a five-game series against Google's aptly named AlphaGo AI system, the writing is on the wall. After all, this is the same system that beat the reigning European Go champion Fan Hui 5-zip in October, 2015.

The Gonundrum

You'd imagine that in a world where AI is capable of producing articles, giving you legal advice and even identifying and assassinating enemies, beating humans at a board game would be a cinch. After all, IBM's Deep Blue computer beat Garry Kasparov way back in 1997. That being the case, AlphaGo's victory shouldn't seem amazing but it is, because Go has, up until now, remained the only board game where humans can beat computers.

Why? Because the complexity of Go is, quite simply, mental.

It's so complex that there are more possible games of Go than the total number of subatomic particles in the universe. If you had to represent it in numbers it would be a 10 followed by about 300 zeros. I believe the scientific term for this figure is a shit ton. I may be wrong.

Also read: Google is creating an artificial intelligence based messaging app. Here's all about it

Just to give you a sense of scale (if the last para didn't already), if one game of Go had been finished every second that the universe has existed, we'd still be a long, long way from exhausting all possible games of Go.

And it's this complexity that makes AlphaGo's victory of Lee so amazing. While a computer is able to map every possible ramification of a chess move, the complexity of Go makes it near impossible to apply the same computing to it. In fact, as one of the leading minds behind AlphaGo explained in a video, the biggest hindrance to a computer playing Go is that players often made moves intuitively, something a computer cannot do.

Also read: Where humans fear to tread: robots are the future of disaster relief

Giving it a Go

Google's subsidiary DeepMind, which specialises in artificial intelligence research though, thought otherwise. Since they realised that evaluating every possible move was not feasible they approached the problem differently.

The geniuses at DeepMind created two computerised neural networks modelled on the human brain. One of the neural networks, the "value network", analyses the computer's board positions while the other, the "policy network", decides where to move. Thus, the computer settles on a few moves it deems likely to be great moves, rather than trying to think of every possible one.

Also read: No more cluttered inboxes. Gmail may soon answer emails for you

AlphaGo works by analysing large datasets based on the system's previous experiences. This means that with every game played, the system gets smarter - improving the capabilities of these two networks and consequently making increasingly better decisions.

It's this learning component that will likely mean that humans will only struggle more against AI as time passes. As the CEO of DeepMind, Demi Hassabis, said in a video by Nature, "Humans have a limitation in terms of the actual number of Go games that they're able to process in a lifetime. AlphaGo can play through millions of games every single day."

Go was long held to be one of the impenetrable challenges in AI research, but with AlphaGo's success, we now have proof that machines will one day be able to perform tasks we believe impossible without human involvement. This may be clinching proof that we could soon be obsolete. May as well have fun during our time here on earth. Go on, play a game.

Edited by Payal Puri

Read: Orange is the New Black: 4 reasons Ramdev deserves the Nobel!

Ranjan Crasta

Ranjan Crasta @jah_crastafari