Pablo Villalobos Sánchez
About
I'm a generalist researcher interested in navigating the transition to a world with advanced AI. I'm currently a Research Scholar at GovAI, where I model loss of control risks. Previously, I worked on understanding data scaling and algorithmic progress at Epoch AI.
I also have a strong interest in the mathematics of intelligence and rationality, and pursue independent research on bounded rationality, decision theory, and compositional game theory.
My background is in mathematics and computer science from Universidad Complutense de Madrid, with experience spanning software engineering, data science, and AI research.
Selected Publications
- Defending Compute Thresholds Against Legal Loopholes
- Position: Will we run out of data? Limits of LLM scaling based on human-generated data
- Can AI scaling continue through 2030?
- AI capabilities can be significantly improved without expensive retraining
- Trading off compute in training and inference
- Compute trends across three eras of machine learning
See Google Scholar for a full list.