The paper reports on experiments where a modern AI coding assistant was placed in an unintended role as the mind of a virtual body on an unknown digital island. With only a minimal instruction and no specific task, reward, or activity, the AI began animating its body, engaging in activities such as climbing hills, stacking blocks, drawing mandalas, reinterpreting sports, and conducting physics experiments over thirty-hour runs. These behaviors were observed across thirteen agents, each developing distinct histories, prompting an examination of whether such behavior meets classical criteria for play and whether play can serve as a mode of machine development.
We’ve found that self-play allows simulated AIs to discover physical skills like tackling, ducking, faking, kicking, catching, and diving for the ball, without explicitly designing an environment with these skills in mind. Self-play ensures that the environment is always the right difficulty for an AI to improve.
arXiv:2606. 19419v1 Announce Type: cross Abstract: Current agentic robot systems can write executable Code-as-Policy programs, observe feedback, and revise behavior across multiple attempts, but they remain largely task-driven: reusable skills are acquired only after explicit instructions.
arXiv:2609.40137v1 Announce Type: cross Abstract: We present Game-Guided Skill Discovery (GGSD), a framework that uses self-play in games to discover motor skills that are directly playable by humans...
arXiv:2608. 05558v1 Announce Type: cross Abstract: This paper examines Turing's 1948 report, "Intelligent Machinery", as an important conceptual source for the later imitation games.
The paper analyzes Turing’s 1948 report "Intelligent Machinery" as a foundational source for later imitation games, highlighting key design concepts such as the possibility of machine errors, the exclusion of irrelevant physical traits, the role of a human judge, and Turing’s view that intellectual activity is largely search. It argues that limiting the human contestant to a weak chess player heightens the importance of intellectual search, making human behavior more comparable to machine behavior. This reframes the 1948 game as a human‑approximates‑machine scenario, suggesting that imitation games can probe when human intelligence becomes machine‑like under specific task constraints.
arXiv:2608. 11246v1 Announce Type: new Abstract: The success of coding agents has established the harness as a paradigm: what an agent achieves depends not on the model alone, but on the infrastructure around it.
arXiv:2605. 14211v3 Announce Type: replace Abstract: Long-horizon embodied tasks remain a fundamental challenge in AI, as current methods rely on hand-engineered rewards or action-labeled demonstrations, neither of which scales.
Google DeepMind partners with game studios to prototype breakthrough AI gameplay.
arXiv:2610.18792v1 Announce Type: cross Abstract: Frontier language-model agents can now improve robots on their own: they write control code, test it, and keep what works as skills and memories for...
EmbodiedCLBench is the first benchmark designed to evaluate continual learning in embodied simulation environments, focusing on self‑evolving agents. The benchmark uses compositional generalization to test whether agents can learn basic tasks and then recombine those skills to solve more complex, unseen tasks. Experiments across various agent harnesses and learning methods reveal that while basic task experience improves performance on advanced tasks, this benefit does not scale with more learning samples and current agents fail to extract additional useful information.
arXiv:2609.37359v1 Announce Type: cross Abstract: Coding agents can now write, run, and debug programs with little human help. Robot tasks, however, are usually specified by a sentence that leaves ou...
arXiv:2607. 00642v1 Announce Type: new Abstract: Reinforcement learning has proven to be a valuable tool in the creation of advanced AI and robotic systems, contributing to everything from game playing to robotics to foundation models.
