Google's Supercomputer AlphaGo Beats World’s Top Player, But Not Without A Struggle

Google’s AI Computer Beats World Go Champ, But Not Without a Struggle

by -

After Lee Sedol defeated AlphaGo on the 5-hour long match on Sunday, all his supporters hoped that the world Go champion would do the same in the Monday match. But the supercomputer refined its moves and went ahead to scoop a 4 -1 win against its human opponent.

Match Pitting World Champ Against Artificial Intelligence

That a computer could beat a world Go title holder is still a wonder. Lee Sedol himself had predicted a 4-1 win against the computer, and many people who have seen Sedol play would have told you before the match that he was going to tackle the machine hands down. A professional Go player who was also a commentator for the match, Michael Redmond, said that he had to see the computer beat the world champion in three consecutive games before he became convinced.

Go, an ancient Eastern game, is said to be more complex than chess. The game involves two players taking turns in arranging either white or black chips on a board with the aim of acquiring the largest territory possible. As easy as the explanation may seem, there are lots of loops and rules a player must meander through to achieve the target and to avoid giving an advantage to their opponent.

Each game between Sedol and AlphaGo was played on a separate day, and lasted between three and five hours. Fan Hui, another top player from Europe, had faced AlphaGo months earlier but never managed to win a single match.

Computers are Game Masters

Computers, it seems, are now the official masters of board games. Go is the latest addition in the long list of games that computers have taken over.

While other board games, including chess, have been won against experts by feeding a set of pre-determined steps to a computer, this is rather impossible with Go, which has too many possibilities that are not easy to summarize this way.

AlphaGo Can Learn and Reason

AlphaGo’s win is attributed to the different technology that allows the computer to reason and make its own decisions based on both immediate and earlier steps that opponents have taken. This achievement is a welcome addition to the growing list of successful and diverse inventions that Alphabet, Google’s holding company, is working on.

To achieve expertise, the AI itself must play many games against competent people. During each game, apart from learning tricks applied by its opponents, the machine also keeps memories of its earlier steps and judges whether they were fruitful or not; if they were good, it keeps a record of them to reuse them when it finds itself in a similar situation next time; if they are not useful, it discards them.

AlphaGo is a Relaxed Gamer

Another advantage that the machine has over its competitors is that it does not experience human weaknesses. Sedol, for instance, seemed to have had much confidence when he was coming to the game but this quickly waned after he was vanquished in the first round.

During the second round, he was visibly tense, and this could have contributed to a few mistakes he committed during the second and third rounds.

In the last two rounds, Sedol showed more stamina and a renewed confidence. This is what must have led to his victory in the fourth round. The loss in the fifth game may have arisen partly from the machine mastering most of Sedol’s moves and partly due to fatigue. However, all the viewers and commentators agree that it was a tough and exciting battle.

Google Deepmind CEO: AlphaGo Learned a lot

During the five games it played with Sedol, the CEO of Google DeepMind, Demis Hassabis, admitted that the computer acquired a lot of lessons.

Hassabis had emphasized long before the match that whether AlphaGo won or lost, it would still be a big victory to humanity. If Sedol won, it would have meant that the company puts more effort into the technology.

However, the win by AlphaGo is not just about games; it brings with it a new perspective into computer development that will add a lot of value to many other areas of human life. Google will not only use these new developments in its voice app development but will also apply it in searching for solutions for climate change and healthcare issues.

Note: The opinions expressed in this article are the author's own and do not necessarily reflect the view of Alvexo on the matter.