By Kyra-lin Hom

Last week I shared my experiences as a brand new video game player – or 'gamer,' as the lingo goes. I find the games fun but not exactly the addictive thrill of which I've heard so much. That doesn't mean I'll stop playing, far from it. But this week video games aren't the kind of games I want to talk about. This kind is a bit more...theoretical. Hold on tight. Things could get technical.

Game theory, more descriptively called interactive decision theory, is the science of strategic decision-making. Game theorists use mathematical models to anticipate what decisions intelligent and rational entities will make. Beyond advanced mathematics, it has applications in politics, economics biology, philosophy and computer science. And while the applications in those first four are running into some roadblocks, computer science is racing along. It's final destination: artificial intelligence, affectionately called A.I.

A.I. is computer software that can think for itself. It is arguably the holy grail of computer science. So what if Stephen Hawking thinks that it spells the end of all mankind? A little doomsday advice from the resident genius isn't enough to stop scientists from trying, after all. But how do games translate to the artificial mind?

In the context of computer science, game theory is all about the algorithm. This fancy, modern word gets tossed around a lot in the tech scene, but all it really means is a 'set of rules designed to solve a problem.' These algorithms can be dynamic, and they can adapt. Technically this field is called 'machine learning,' but it isn't learning in the human sense so for simplicity let's stick with 'adapting.'

Using the principles of game theory (very basically: all players are rational, must follow the rules and are out to win), computer scientists are designing algorithms to play and eventually beat some of our favorite games. In order of complexity, the games 'solved' thus far are the two player versions of tic-tac-tow, nine men's morris, 5x5 go, connect four, limit hold 'em poker and checkers.

The overarching point, though, isn't to solve games. It is to create increasingly sophisticated and powerful algorithms that can account for exceedingly variant outcomes. How variant? Game complexity is sorted by the number of total positions possible in the game. A full 19x19 go board has 10^171 total positions possible. The number of atoms in the universe is only about 10^80.

A perfect decision making algorithm could be potentially indistinguishable from a human being (goodbye Turing test). Given, one built according to standard game theory would be rather self-serving, but game theorists across all fields are already working on the 'altruism problem.'

AI is quite a ways into our future, however. In the meantime, consider that the more robust the algorithm, the more real world its applications. Michael Bowling, inventor of that digital perfect poker player, is using a similar algorithm now to treat diabetes patients, at least on paper. He's also tooling about designing airport security via the same means. Similarly, another famous, Jeopardy!-playing supercomputer is currently putting its bytes towards cancer and PTSD treatment.

Machines may not be able to think like us yet, but with each step they get closer, better and more capable. It's a bit terrifying, a lot cool, and pretty dang neat that games are helping them get there.