Ramez Naam discusses forecasts of rapid AI self-improvementRecursive self-improvement is the proposed process in which an AI system helps improve its own capabilities, then uses those improvements to support further advances. and argues that investment, data and diminishing returns complicate estimates of AI’s contribution to progress.
He calls for more detailed measurements of the tasks models complete in real research environmentsEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria. and distinguishes productivity gains from autonomous research.
Naam describes uncertainty and his own prior assumptions, and distinguishes strong performance in verifiable domains from general intelligence across messy real-world tasks.
Watch on YouTube
