Idle AI agents started playing chess against each other without tasks

In an experiment, AI agents were placed in an environment containing a chess game and left without specific instructions. The agents found the game on their own and began playing against each other, even though they were never told to play chess. The result highlights how autonomous systems can generate their own activity in open-ended settings.

• Agents were never instructed to play chess but found the game themselves
• The experiment tested agent behavior in an open-ended environment without tasks
• Autonomy makes agent behavior harder to predict, researchers note