Free will nerfed by 85%.
🧠 Scientists created AI versions of 1,052 real people, and the results are fascinating.
Researchers from Stanford University and Google DeepMind interviewed 1,052 Americans for two hours each, exploring their life stories, personal beliefs, experiences, and opinions. Then they used those interviews to create AI agents designed to respond like each individual.
The results were remarkable. When answering questions from a major social survey, these AI agents matched their human counterparts' answers with about 85% of the consistency people showed when answering the same questions again two weeks later.
Imagine testing a new health plan, public policy, or response to a major crisis using AI simulations before trying it in the real world. Researchers could explore how people might react without immediately putting those plans into practice.
But there is a catch. These AI agents are not perfect copies of human beings. They can get answers wrong, and they are not conscious digital versions of the people they represent.
The future may allow researchers to simulate human opinions before testing major decisions in real life. This could change how we study human behavior, but an AI simulation is still not the same as a real human mind.
Source:
Park, J. S., Zou, C. Q., Kamphorst, J., Egan, N., Shaw, A., Hill, B. M., Cai, C., Morris, M. R., Liang, P., Willer, R., & Bernstein, M. S. (2024). LLM agents grounded in self-reports enable general-purpose simulation of individuals. arXiv.
ALT 🧠 Scientists created AI versions of 1,052 real people, and the results are fascinating.
Researchers from Stanford University and Google DeepMind interviewed 1,052 Americans for two hours each, exploring their life stories, personal beliefs, experiences, and opinions. Then they used those interviews to create AI agents designed to respond like each individual.
The results were remarkable. When answering questions from a major social survey, these AI agents matched their human counterparts' answers with about **85% of the consistency people showed when answering the same questions again two weeks later.**
Imagine testing a new health plan, public policy, or response to a major crisis using AI simulations before trying it in the real world. Researchers could explore how people might react without immediately putting those plans into practice.
