Daily curated AI insights you can't miss.
OpenAI's Chief Research Officer Mark Chen discusses why scaling laws remain central to frontier progress and why pre-training is not dead.
Chen explains the reasoning bet behind o1 and how compute allocation is used to prioritize research directions.
The evals crisis is addressed, with Chen noting how benchmarks can be misleading and why better evaluations are needed.
The podcast episode features Mark Chen, Chief Research Officer at OpenAI, discussing core AI research topics while cooking.
The conversation addresses the ongoing relevance of scaling laws and pre-training, countering claims that progress in this area has stalled.
Chen explains OpenAI‘s strategic bet on reasoning, specifically through models like o1, and describes how compute allocation is used to prioritize high-impact research directions.
He also details the challenges of building effective evaluations (evals) to measure model capability and avoid benchmark overfitting.
