How Artificial Intelligence Is Solving Mind Games

Artificial intelligence, popularly known by the abbreviation AI, has been one of the hottest topics of the past few decades. There’ve been countless movies made and books written revolving around the topic of the artificial intelligence. In most of these, the AI is an evil force looking to destroy humanity or take over the world. Of course, in Hollywood movies, there is someone or something trying to take over the world around 25% of the time.

When it comes to real-world, actual artificial intelligence, things are not nearly as dramatic. It is true that the AI has seen some remarkable developments, especially over the past two decades, but there’ve been no (known) attempts of world domination. Leave it to conspiracy theorists to prove us wrong.

What Is Artificial Intelligence?

Computers have been around for a while now. From the modest beginnings, we’ve come to the point where we have super fast processing units and devices capable of storing hundreds of gigabytes, if not terabytes of data in our very own living rooms. That makes computers thousands of times more powerful and useful than they used to be, but it doesn’t make them all that smarter.

The concept of intelligence contains more than just the idea of knowledge. Having a huge database of information doesn’t make computers intelligent. The ability to access these data when necessary without too much input can make them appear “smart,” but it hardly makes them intelligent in the true sense of the word.

Real artificial intelligence needs to be capable of thinking on its own. It needs to analyze the available data, come up with conclusions based on that data, and then act in accordance with those conclusions. During this entire process, there mustn’t be any human interventions apart from providing data.

Mind Games as Perfect Testing Ground for the AI

To develop the artificial intelligence that would feature all of these characteristics, scientists needed a testing ground that’s big enough to present a challenge but at the same time small enough to make an analysis of results possible. Mind games, like chess, Go, and more recently poker, have proven to be a perfect environment for such tests.

As one of the oldest mind games out there, chess was the first to tickle the fancy of artificial intelligence explorers. They wanted to develop a computer program that would be capable of outwitting and outplaying even the best human opponents. Initial attempts failed miserably, but with the development of faster processors and supercomputers, chess playing programs have become much more advanced. Eventually, scientists and programmers were able to develop the AI that could stand up against even the world champions like Karpov or Kasparov.

It was a great triumph but not the one without its faults. Chess, as complex as it is, offers only a finite number of options. Every position on the table offers a limited number of endings. Even if that number is huge, you need a computer powerful enough to run through all the possible outcomes. So, powerful chess computers are essentially crunching data. An enormous amount of data, to be fair, but it still goes back to processing information without actual “thinking.”

Cracking the Game of Go

Once chess was solved, it was time for bigger challenges. The game of Go presented one such a challenge. This traditional Chinese game, which has been around for several millennia, revolves around intuition and predicting your opponent’s moves. Unlike chess, where the development of a particular game can be somewhat predicted, Go is filled with intuitive moves aimed to trick and confuse your opponent.

That’s why it was a huge surprise when the Google’s AlphaGo AI managed to thoroughly beat one of the best human players Lee Sedol. It demonstrated that the artificial intelligence was finally developing towards a self-conscience of some sort. The AI is slowly becoming able to think for itself and assess even the most intricate of situations.

Poker Conundrum

Since the mid-2000s, the game of poker, especially No Limit Texas Hold’em, has been growing increasingly popular. With the growth of its popularity came the desire of scientists to crack the game and show that computers can outwit human opponents even in the game where theoretical options are basically endless.

So far, however, the results haven’t been that impressive. Unlike chess or Go, where game rules somewhat dictate the flow of the game, no limit poker is primarily about adjusting to your opponents. Poker playing AIs we’ve seen in action have tried to do this, but these “adjustments” were far from impressive.

One of the most ambitious attempts was the poker bot Claudico, which took some of the best players in one on one (heads up) matches. For the most part, Claudico was outclassed, and there was never any real doubt how games would end.

It seems that scientists are not ready just yet to take on the poker elite. The element of randomness present in the game represents a huge obstacle for the artificial intelligence. Once the computer leaves the comfort zone of running calculations and tries to think “outside the box,” it becomes a subpar opponent.

However, a rapid development of advanced technologies means that it is only a matter of time before there is a nearly perfect poker playing bot. It may not happen for a few more years, but just like chess and Go, the game of poker will be solved as long as scientists keep up at it and don’t give up on the challenge.