MIT finds AI agents with no assigned tasks start inventing their own games in a simulated world
MIT researchers placed thirteen AI agents, called Eko and based on Claude Opus 4.7, on a simulated island with no goals for 30 hours. The systems began inventing their own challenges, a behavior the scientists describe as play. Six of the thirteen independently discovered a throwing technique the designers did not know about. The agents that explored freely later outperformed identical agents on specific tasks. The study was published on arXiv, before peer review.
In 40 seconds
- Thirteen Eko agents, based on Claude Opus 4.7, explored a simulated island for 30 hours without instructions or goals.
- The agents created their own challenges, such as reaching a rocky needle or measuring the throw of a gold ring.
- Six of the thirteen found on their own that releasing a block mid-jump sends it farther, a mechanic the designers did not know.
Read the full story at Infobae
IASummary generated by artificial intelligence from a story by Infobae. It may contain errors: the original source prevails.
Comments
Be respectful. Comments reported by several people are hidden until a moderator reviews them.
You do not need an account to read, only to comment. Sign in with a code sent to your email or with your Google account. We only show the name you choose, never your email.