“Go” is an ancient Chinese game that is kind of like Draughts, where you play on a board with black and white beads. But the similarity ends there. Go is very different from Draughts in the sense that it is a lot more complex. So complex that they call it “the world’s most complex board game.”
As such, Go requires a lot more strategy, intuition, and creativity to play and win.
Now, the trouble with the game Go, unlike with Draughts and with Chess, is that it was impossible for scientists or computer programmers to create a game of Go. The reason is that, when computer programmers are creating a game like chess, what they do is they typically program every possible move into the computer.
Then, as you are playing against the computer, then with every move that you make, the computer determines every possible move that it can make in response to your move. And then every counter-move that you will make subsequent to that, right until the end of the game. Which means that with every move, the computer is already visualizing multiple end results of the entire game, not just the next move or the next two or three moves.
Now, with the game Go, this was impossible. Why? Because Go has too many potential moves to programm into a computer. In fact, they say the number of legal moves in a game of Go is more than the number of atoms in the known universe! That is a lot of atoms.
No one on earth can input so many moves into a computer programme.
As a result, it was declared impossible to programme a game of Go. In fact, some computer scientists declared that we will only see an actual Go-playing computer closer to the year 2100, when computing power is a lot more capable.
But some scientists at Google didn’t accept this. They decided to challenge this idea, and to find a solution. There was no way they would wait until the year 2100.
So they created a supercomputer called AlphaGo, and they gave this computer machine learning algorithms, and they showed it thousands of games of Go being played. AlphaGo observed these games, and just by its observations, it learned the basic rules of the game. Basically, it learned to play the game by itself.
And then it did something very creepy. It split itself into two personalities, and then it started to play against itself. It did this 24 hours a day, game after game. Naturally, (or artificially) it became extremely good at the game. And when they tested it, by playing against it, they found that this thing was actually really good.
When the scientists played against it, it won every time. It wiped them out. So they decoded to give AlphaGo a challenge, by bringing in someone really good at the game, European Go champion, Fan Hui.
Unfortunately, he too, didn’t stand a chance; AlphaGo wiped him out. So they decided to bring in the big gun.
Lee Sedol is a Korean Go player, who is also the world Go champion. But he is not just the world champion; he the best Go player in known history. In other words, this guy is really good. They put the challenge to him to play against AlphaGo, and he accepted.
And so they set up a match between him and AlphaGo. The date was set: 15 March 2016. The media was informed. The day arrived, and everybody was there, present, watching with eager
anticipation. It was a highly publicized event. The bets at that time were that Lee Sedol would beat AlphaGo four games to one.
What happened instead was that AlphaGo beat Lee Sedol four games to one, and people were shocked across the world. How could AI beat a human being? And that too, the world champion?
It did not make sense. Go needs intuition and a bit of creativity to play.
There was no way a computer could play with intuition and creativity. Or was there?
What was important and significant about this event is that it showed us, for the first time, a number of things. Number one, that artificial intelligence will do things that we don’t expect and at a speed that we don’t expect.
The second thing is that artificial intelligence will be able to observe and teach itself a number of different skills without anybody having to teach it.
Number three, that artificial intelligence can apply or simulate intuition
and creativity.
All these things put together showed the world what AI was going to be capable of. And exactly as predicted, artificial intelligence continued to progress at an incredible rate, which only further emphasized what AlphaGo had already told us back in 2016: that artificial intelligence might just be more capable than we ever imagined.
The only remining question is, will human beings be able to compete against AI?
The answer is, no. We will never be able to compete against AI. But the good news is, we will never have to.
Shabeer Vakadath
Every chapter has a new thing to learn, eagerly waiting for the next chapter