Research Fellowship, Evolution of AI Capabilities at the Frontier and Thresholds of Human Performance

Paris, France

• In this role, you'll extend the Epoch Capabilities Index using Bayesian methods to map AI capability thresholds against human performance.

• Develop a multidimensional Rosetta Stone framework integrating human performance benchmarks.

• Identify and compare current AI capabilities against different levels of human expertise.

• Build statistical projections forecasting when AI systems surpass expert performance.

• Apply Bayesian modelling (Stan, PyMC, Pyro) and data science analysis throughout.

The General Purpose AI Policy Lab is a French organization specializing in security and international coordination issues raised by frontier general-purpose AI models.