DeepMind researchers have developed AI agents capable of playing virtual football.
According to the engineers, the AI agents learned the game almost from scratch. First, virtual robots learned to walk, and then run and kick the ball.
At each new stage, artificial intelligence systems were shown videos with real football players. This allowed them to learn the basics of the game, imitate the movements of professional athletes while participating in real high-level events.
As soon as the robot learned to play alone, it was paired with another AI agent. As the skills of virtual soccer players improved, the number of players increased.
As a result, the researchers formed small groups competing against each other.
“The result was a team of coordinated humanoid football players exhibiting complex behavior at different scales, quantified by a range of analyzes and statistics, including those used in real-life sports analytics,” the study says.
According to the engineers, each football player robot makes a decision during the game on its own. However, they recognized some simplifications in the rules. For example, the system does not take into account fouls, and an invisible border is set around the field, preventing the ball from going out of its borders.
The researchers also noted that it took a long time to train the agents. This could make it harder to scale the technology to physical robots, they added.
Recall that in July, DeepMind’s AlphaFold algorithm predicted almost all proteins known to science.
In August 2021, the laboratory presented a universal architecture for creating artificial intelligence.
In June, scientists at DeepMind said that reinforcement learning was enough to achieve artificial general intelligence.
Subscribe to Cryplogger news in Telegram: Cryplogger AI – all the news from the world of AI!
Found a mistake in the text? Select it and press CTRL+ENTER
DeepMind researchers have developed AI agents capable of playing virtual football.
According to the engineers, the AI agents learned the game almost from scratch. First, virtual robots learned to walk, and then run and kick the ball.
At each new stage, artificial intelligence systems were shown videos with real football players. This allowed them to learn the basics of the game, imitate the movements of professional athletes while participating in real high-level events.
As soon as the robot learned to play alone, it was paired with another AI agent. As the skills of virtual soccer players improved, the number of players increased.
As a result, the researchers formed small groups competing against each other.
“The result was a team of coordinated humanoid football players exhibiting complex behavior at different scales, quantified by a range of analyzes and statistics, including those used in real-life sports analytics,” the study says.
According to the engineers, each football player robot makes a decision during the game on its own. However, they recognized some simplifications in the rules. For example, the system does not take into account fouls, and an invisible border is set around the field, preventing the ball from going out of its borders.
The researchers also noted that it took a long time to train the agents. This could make it harder to scale the technology to physical robots, they added.
Recall that in July, DeepMind’s AlphaFold algorithm predicted almost all proteins known to science.
In August 2021, the laboratory presented a universal architecture for creating artificial intelligence.
In June, scientists at DeepMind said that reinforcement learning was enough to achieve artificial general intelligence.
Subscribe to Cryplogger news in Telegram: Cryplogger AI – all the news from the world of AI!
Found a mistake in the text? Select it and press CTRL+ENTER
