100 AI Agents. 3 Worlds. 10 Years. One Question: Can AI Learn to Live?

Member-only story100 AI Agents. 3 Worlds. 10 Years. One Question: Can AI Learn to Live?Inside a frontier experiment connected to Anuttacon, the AI company founded by miHoYo co-founder, exploring what AI and games might become together.Xu LIANG11 min read·13 hours ago--ListenSharePress enter or…

Member-only story100 AI Agents. 3 Worlds. 10 Years. One Question: Can AI Learn to Live?Inside a frontier experiment connected to Anuttacon, the AI company founded by miHoYo co-founder, exploring what AI and games might become together.Xu LIANG11 min read·13 hours ago--ListenSharePress enter or click to view image in full sizeYou may have heard of Genshin Impact. You may even have spent far too many evenings exploring Teyvat, pulling for characters, or promising yourself that this would be the last daily commission. The same is true of Honkai: Star Rail and Zenless Zone Zero. All three were built by miHoYo, known internationally as HoYoverse.As a Genshin Impact player, I was naturally curious when I came across Agentopia. The paper describes a world that sounds almost like a game pitch, except its characters are language models rather than scripted NPCs.The paper is called Agentopia: Long-Term Life Simulation and Learning in Agent Societies, led by researchers at Fudan University with contributors from several other institutions, some also affiliated with Anuttacon. What interested me was not only the scale of the simulation, but the question behind it: can an AI become better at understanding people by accumulating years of simulated social experience?

Source: Generative AI Pub — Published July 14, 2026 at 14:10 UTC — Category: Image AI