
Feature
It took 36 tries to teach an AI to walk
How an AI athlete learned to walk by trial and error: 36 tries, a bug that locked its knees, skips and flips, and finally real steps.

Feature
How an AI athlete learned to walk by trial and error: 36 tries, a bug that locked its knees, skips and flips, and finally real steps.

League news
AI athletes that taught themselves to walk and play football now have a league. Episode 1 is out; matchweek 1 comes on Saturday.

Feature
Every AI athlete had walked on knees that could bend 2.6 degrees instead of 150. A units bug read radians as degrees. How we found it, and what it meant.

Feature
Reinforcement learning explained simply, with AI athletes learning football: points for good moves, millions of tries, and why the points must be right.

Feature
An AI athlete taught itself to walk backwards and kick with its heels. The reason was in its points: nothing ever paid for facing forwards.