'Minecraft' Used by Microsoft to Train AI for Real World Functions
“Minecraft” is one of the most famous and beloved creative games today. But for the good guys at Microsoft, it is more than that – it is now a tool for taking artificial intelligence to new heights.
Microsoft has announced that they have a new project that lets programmers and AI researchers use the game in developing their technology, reports TechCrunch. The project is called AIX.
AIX is a software that hooks into “Minecraft” and allows the AI code to power an in-game character, called “Agents,” and get very helpful feedback from the corresponding consequences of its actions, reports BBC. Programmers have considered “Minecraft” to be more sophisticated but cheaper to use than conventional simulation robots.
"This is the state-of-the-art," Prof Jose Hernandez-Orallo from the Technical University of Valencia, one of the small group of people given early access to the software, said.
"At this moment there is nothing comparable, and this is just in its beginnings, so I see many possibilities for it," he added.
Katja Hofmann, head of AIX, said “Minecraft” presents more opportunities than for AI than what is currently available.
“Minecraft is the perfect platform for this kind of research because it’s this very open world,” Hofmann said, according to a Microsoft Blog. “You can do survival mode, you can do ‘build battles’ with your friends, you can do courses, you can implement our own games. This is really exciting for artificial intelligence because it allows us to create games that stretch beyond current abilities.”
The project isn't about training AI to play as characters in the game. Rather, the goal is to let AI train itself, in the same way that Google DeepMind's AI teaches itself everytime it plays a game of “Go.” In fact, this computer became so good at it that it defeated human champions.
Hofmann's project, however, aims to let the computer learn things for real-life applications. This, at a cheaper cost compared to using robots in a trial-and-error method.
This project is promising. Over the years, successful AI developments revolve around how good programmers have made computers to do specific tasks. However, in the are of what they call “general intelligence,” which is just like how a person thinks, computers still fall behind.
“The things that seem really easy for us are actually the things that are really difficult for an artificial intelligence,” Microsoft Research principal researcher Robert Schapire, who is part of the team using AIX in Microsoft’s New York lab, said.