There are countless films from the sci-fi genre where artificial intelligence begins to get the better of humans. In these films it generally has pretty dire consequences that require immediate human interference. Sometimes it ends in the humans and robots getting along famously, but more often it ends in enormous explosions. Whether you’re aware of these films or not, they might be closer to the truth than you would first think, though hopefully not with the explosions. These are some of the areas where artificial intelligence can beat us humans at our own game.
Chess
One of the first artificial intelligence programs to really get the better of humans was The Deep Blue. This computer was designed specifically to compete in chess and has been outperforming humans since 1997. For over a decade, computers had already been able to beat humans at chess with a degree of success, but The Deep Blue was (and still is) brilliant at it. It beat the World Chess Champion Garry Kasparov in 1997 and hasn’t looked back since. Although it hasn’t come up against the reigning champion Magnus Carlsen yet, the statistics that it has kept up so far suggest it wouldn’t have too much trouble with him. We’ll have to wait and see what happens in that match!
Poker
One of the areas that has made huge progress over the past few years is within the realm of poker online. In this sector, artificial intelligence systems have been able to beat professional poker players at their own game. The program that’s had the most success is called Deepstack. Whilst beating masters at chess is impressive, it is actually harder to beat professionals at poker reliably. This is because chess is what is known as a game of perfect information; that is, you know where all of the pieces are on the board, and you know how they can move. With poker, a lot of the information is hidden; even if you spent all day working out the current hand, you still wouldn’t know for sure which cards your opponent held. Artificial intelligence has struggled for a long time with these games of imperfect information, but the team at Deepstack have been working to improve results and they have, remarkably. The program is capable of beating professional poker players thanks to its ability to react quickly, using only limited information. It’s as close to human thinking as anyone has got within the poker artificial intelligence realm, meaning that in a test of more than 40,000 games against the professionals, Deepstack beat all of them, bar one.
Driving
Caption: For now, humans have the edge when it comes to driving
The era of the self-driving vehicle is coming and although there have been some pretty high profile mishaps with driverless cars, there have been a great deal more successes. It was predicted that we’d see millions of driverless cars on the roads by now, by last year in fact, though they haven’t yet arrived. Whilst under test conditions driverless cars outperform humans, when it comes to being let loose on actual roads, us humans still have the edge. This is another case of imperfect information being a stumbling block for artificial intelligence. Whilst at the poker table it’s okay to lose some of the time, as long as you win most of the time, on the roads it simply isn’t good enough. If you crash once, you could kill somebody. The subtle cues that us humans give to one another – a widening of the eyes, for example, to indicate that the other driver can move forwards – is the sort of thing that artificial intelligence can’t yet rely on. So, this one is a bit of a trick, they’ve not got it yet, but when they do it will be revolutionary.
