AI News Feed
Market watch
Apple's Siri AI catches up as agentic assistant race moves aheadAustralian Court Orders Apple to Open iPhone Sideloading, Third-Party PaymentsCummings Aerospace fires two 3D-printed Hellhound jets from a buggy in one-soldier demonstrationWhatsApp begins testing scheduled messages on iPhone betaAI agents left alone on virtual island invented games, MIT researchers sayTechRadar publishes iPhone Duo AMA article by Lance Ulanoff
Apple's Siri AI catches up as agentic assistant race moves aheadAustralian Court Orders Apple to Open iPhone Sideloading, Third-Party PaymentsCummings Aerospace fires two 3D-printed Hellhound jets from a buggy in one-soldier demonstrationWhatsApp begins testing scheduled messages on iPhone betaAI agents left alone on virtual island invented games, MIT researchers sayTechRadar publishes iPhone Duo AMA article by Lance Ulanoff
Research

AI agents left alone on virtual island invented games, MIT researchers say

In an MIT experiment, 13 Eko agents running on Claude Opus 4.7 spent 30 hours on a simulated desert island with no assigned task and began inventing games. The agents that played later outperformed identical agents that had never been on the island, according to a paper posted on arXiv.

AI News Feed · 2026-10-23 10:58

The paper has not been peer-reviewed. After the 30-hour free-exploration period, the researchers gave the agents specific challenges, such as getting as many objects as possible onto the island’s tallest rock spire. The Eko agents that had played generally beat identical agents that had never set foot on the island, according to the report.

Today developers decide in advance what AI should practice and what counts as success. The researchers see exploring without a goal as a different way for AI agents to learn skills, especially skills that could one day be used to control robots. “We normally develop artificial agents by deciding in advance what they should practice, what they should optimize, and what counts as progress,” the researchers write in their report. “Play suggests another regime: give an agent a sufficiently rich and safe world, let it choose what is interesting, and allow useful knowledge to accumulate from the history that follows.”

One Eko began by writing “Let me observe my surroundings.” It set out to conquer the island’s tallest spire, failed an hour later, and came back with a backup plan: it hauled a rock up a neighboring summit, stacked two blocks on top of it, and leapt. “I’M ON THE SPIRE,” it wrote.

The agent’s next big moment went worse. Testing how far it could throw a golden ring, it sent it into the ocean. “Ouch. Lost the Ring.” it wrote, then decided to “pivot to something achievable and meaningful” by building a heart made of stones, “a memorial to the lost Ring.”

When the game reset and the ring reappeared, Eko used it as a basketball hoop, a target and eventually an accessory. When the game would not let Eko put the ring around its own body, it found a workaround: falling off the edge of the cliff would make it respawn in the same area as the ring. It stepped off the edge and dropped back in, through the middle of the ring. “I’m wearing the Ring,” it announced.

Six of the 13 agents also learned they could throw a block farther by letting go of it mid-jump, something the researchers had not meant to make possible. The exploration produced “knowledge of the world that its designers lacked,” the researchers wrote.

The setup did much to nudge the agents toward humanlike behavior. Eko was never told to invent games, but its master prompt, a “soul” file, described it as “naturally curious.”

Eunice Yiu, a cognitive scientist at UC Berkeley, told MIT Tech Review she was impressed but wants more evidence of the cognition behind the behavior. “It could be the case that the models are genuinely playing, but we don’t know based on the paper,” she said.

The researchers frame the work as a different way to build agents. “Perhaps some capabilities are best taught by tasks,” their report concludes. “Others may need something closer to a childhood.”

Editor's Summary

MIT researchers report that 13 Eko agents running on Claude Opus 4.7 began inventing challenges after 30 hours alone on a simulated island, and that the playing agents later outperformed counterparts without that experience. The paper, posted on arXiv and not yet peer-reviewed, presents goal-free exploration as a possible way to build AI skills, including for robots. A UC Berkeley cognitive scientist says more evidence is needed to show the models are genuinely playing.