The OpenAI Podcast: Why Tejal Patwardhan Stopped Underestimating AI

Tejal Patwardhan explains why the biggest mistake in AI is underestimating what models can already do. In this condensed summary of The OpenAI Podcast, host Andrew Mayne talks with Patwardhan about capability overhang, benchmarking, reasoning models, AI safety, and why narrow evals can miss real-world progress. Hear how OpenAI’s research team moved from academic tests to harder evaluations involving codebases, science tasks, cybersecurity, and multimodal systems, plus why a surprising GPQA result and a CTF breakout moment changed their view of model capabilities. You’ll come away with a clearer understanding of how AI benchmarking, alignment, and leadership decisions shape what comes next in artificial intelligence and machine learning. Listen now to get the key ideas in minutes.