It even cheated: for the first time, an AI robot was able to outperform humans “physically”

Researchers from ETH Zurich have created an artificial intelligence robot called CyberRunner that outperformed humans in the popular game “Maze”

It became known to agencyBloomberg.

This is the first time that AI has defeated humans when using its “physical” skills. Previously, AI demonstrated only its intellectual abilities by beating humans in virtual chess, Go, or cards. During the experiments, the neural network-based robot used two handles to control the playing surface, using fine motor skills and spatial thinking. The game itself required real-time strategic thinking, quick decisions, and precise actions.

The CyberRunner robot from ETH Zurich managed to outwit people during the game “Maze.”

The AI model used by the scientists belongs to a special type of machine learning, when a neural network learns to behave in space using trial and error. Researchers believe that thousands of such artificial intelligence systems will soon appear, conducting joint experiments, communicating, and sharing best practices.

The capabilities demonstrated by CyberRunner are a new level in the development of robotics, researchers say. The system can think, learn, and evolve, performing physical tasks that were previously thought to be achievable only with the help of human intelligence.

CyberRunner learns from its own experience, through a camera looking at the maze. He was even able to “cheat” by skipping parts of the maze. The researchers had to intervene and directly instruct him not to use such tricks.

Source focus
You might also like
Comments
Loading...

This website uses cookies to improve your experience. We'll assume you're ok with this, but you can opt-out if you wish. Accept Read More