Scaling Theory #18 – James Evans: Science in the Age of AI
May 7, 2025
In this engaging discussion, James Evans, a Professor of Sociology and Computational Science at the University of Chicago, shares insights on how AI is transforming research methodologies. He explores the potential and pitfalls of using large language models to simulate human behavior, highlighting their impact on social dynamics and policy-making. The conversation delves into the future of research collaborations, emphasizing the importance of small teams and interdisciplinary innovation. Evans also reflects on the philosophical implications of viewing humanity as part of a grand experiment.
AI Snips
Chapters
Books
Transcript
Episode notes
AI Simulations Fit Data-Scarce Fields
- AI simulations will first be common where data is fundamentally unavailable, such as future and hidden pasts.
- They help explore existential risks and populations not well captured by existing data.
LLM Agents Fit Complex Systems
- Large language model agents can simulate social systems where first principles models lack are insufficient.
- Social sciences especially benefit because interactions, not particles, create complex phenomena.
Sensory Deficits Limit AI Simulation
- Lack of sensory input limits AI's understanding and simulation quality.
- More sensory data must be integrated to avoid hallucinations and degradation in simulated outputs.