Michael Aaron

This talk addresses the challenges of AI evaluations, which are currently scattered, quickly become outdated, and often lack transparency and verifiability. The speakers propose solutions to democratize the evaluation process, enabling a broader community to contribute to and benefit from robust AI benchmarking. Their work aims to foster more equitable AI development by allowing diverse expertise to shape evaluation standards.