Deepmind taught Jeenets to play football

DeepMind laboratory researchers have developed AI-agents who are able to play virtual football.

At every new stage, artificial intelligence systems were shown by a video with real players. This allowed them to learn the basics of the game, to imitate the movements of professional athletes while participating in real events of a high level.

As soon as the robot learned to play alone, he was set in a couple of other Jeanie. As virtual football players’ skills improve, the number of players increased.

As a result, researchers formed small groups competing against each other.

“The result was a team of coordinated humanoid football players showing complex behavior on different scale, quantitatively determined by a number of analyzes and statistical data, including used in real sports analytics,” the study says.

According to engineers, every tributist robot makes a decision during the game on his own. However, they recognized some simplifications in the rules. For example, the system does not take into account the fouls, and an invisible border is installed around the field Обеспечение ликвидности that prevents the ball output beyond its limits.

The researchers also noted that it took a lot of time to teach agents. This can complicate the scaling of technology to physical robots, they added.

Recall that in July, the Alphafold algorithm DeepMind predicted almost all the proteins known to science.

In August 2021, the laboratory introduced universal architecture to create artificial intelligence.

In June, scientists from DeepMind said that to achieve general artificial intelligence, there is enough training with reinforcement.

Subscribe to FORKLOG news at Telegram: Forklog AI – all news from the world of AI!