Human v Computer : Games

When I was at school, we had a chess team.  It was a little like the cricket team in that those of us that could play were in it – the “pool of talent” didn’t run that deep.

One time we went to play in a tournament against other schools and all I really remember is the terrifying drive there.  The teacher who took us was known for not being a great driver.  I still see him nowadays.  He looked about 60 years old 20 ago.  He is now nearly 60 and looks no different.

One thing he suggested we do was watch a chess match between Grand Master Garry Kasparov and Nigel Short.  I remember tuning in to Channel 4 every night for a few, but chess isn’t really a spectator sport.

Kasparov was actually the first world champion to lose a match to a computer under standard time controls, when he lost to the IBM supercomputer Deep Blue in a highly publicised match.  This was a big deal at the time, and it was probably my first time hearing about artificial intelligence.

Nowadays there is obviously lots of talk of what robots will replace humans in, from our work to our partners.  Advances in machine learning mean that computers are more and more becoming viable alternatives to humans.

That is, rather ironically, other than when it comes to computer games.

Song Byung-gu, a 29-year-old professional StarCraft player, defeated four different bots in succession in less than half an hour at a man-machine face-off in Seoul that highlighted limitations in the power of AI. Contests between two human players of the sci-fi game can last for up to an hour. The bots were capable of carrying out up to 19,000 actions a minute, while even the best human players, including Mr Song, can manage only a few hundred, but they were nonetheless roundly defeated.

Loads of companies including Facebook and DeepMind took part, lending their tech for an ass whooping.  However, it’s thought that there machines failed because the game, StarCraft, requires players to use their memory, devise their strategy, and plan ahead simultaneously, with the map only partially observed. Apart from beating the opponent, a rival alien race, the player must also gather resources and build structures. Decisions made early in the game may only pay off much later.  Apparently.

In other words, there was simply too much the AI didn’t know to plan ahead.  With so many options, it was being indecisive.  I sometimes know how it feels.

I just liked the fact that humans have been computers at their own games.

Leave A Comment?