American gamers will train military robots to "work in a swarm"

The Department of Artificial Intelligence at the University of Buffalo (USA) received a grant of $ 316, 000 to conduct experiments to improve the technique of reading the brain waves and eye movements of gamers. Among the experimental will be only real-time strategy specialists, such as the games of the Starcraft series, since there you need to manage large numbers of small units to solve various problems. The source of funding is DARPA, which sees this as the key to training real robots operating in a swarm.

What exactly to look for in the brains of gamers, neither scientists nor the military do not yet know. It is planned that in addition to traditional games, a special simulator will be created, as well as many tests and puzzles. The main task is to figure out how the player who controls the army of drones analyzes the situation and searches for solutions. These observations will then form the basis of algorithms for training real robots.

To date, the Pentagon has already recognized that a swarm of a million primitive robots for $ 1 is more efficient and convenient than one powerful robot for $ 1 million. But this also creates a problem - such small and simple mechanisms are forced to rely on commands from the outside, they simply do not space for a full-fledged intelligent system for independent work. Therefore, with a normal explosion and smoke on the battlefield, the scattered flock of drones becomes helpless.

The task of DARPA is precisely to develop common simple rules of behavior for robots so that they do not get lost in difficult situations. So that they can separate and collect again, as it should be for a full-fledged structure of the "swarm" type. Perhaps the skills of gamers who are accustomed to controlling dozens and hundreds of units will really help with this.