Harvard & MIT Create 8.3B AI Agents to Simulate Entire Human Behavior
新智元 · wechat · 2026-08-09
A team led by Harvard and MIT PhDs, with contributors from OpenAI and DeepMind, released MatrAIx. The project built Persona-8B, a database of 8.3 billion AI agents mirroring the global population, endowing each with a logically consistent persona across 1,290 dimensions (background, psychology, skills, etc.).
Key Highlights
- Omni-channel Simulation: Agents can be deployed into surveys, AI chats, web browsing, and app interactions (manipulating macOS/iOS) to complete over 1,000 evaluation tasks.
- Near-Human Consistency: In controlled experiments, agent behavior matched assigned personas 91.5% of the time. Claude Opus 4.8's evaluation error was within 1 point of human experts 93.8% of the time.
- Disrupting User Research: Traditional time-consuming user research and A/B testing can now be instantly compressed into a silicon-based sandbox.
However, this raises a philosophical concern: using LLMs to test LLM-driven virtual humans might erase the irrational obsessions and "black swan" anomalies of real humanity, making the future digital world smooth but hollow.
More from AGI Musings
- AI Could Create Unprecedented Wealth But Make Millions Feel Economically Worthless — VraserX · 2026-08-09
- AI Safety Policy Should Be Physics: Controlling Reachable States Over Actions — pstryder · 2026-08-09
- Gwern's Essay: Building Personalized 'Guardian Angel' LLMs for Productivity and Cognitive Security — morgymcg · 2026-08-09
- Breaking the Supervised Learning Curse: Models Must Learn 'In the Wild' — tensorqt · 2026-08-09
- AI Billionaires Pledge to Give Away Fortunes: A Multi-Billion-Dollar Pinky Promise? — Wired AI · 2026-08-09
- China Builds Billion-Agent Society Simulator with Terrifying Predictive Power — pbaylies · 2026-08-09