Sam Altman Returns — Lex Fridman Podcast #419

The March 2024 Lex Fridman conversation with Sam Altman (#419, 3hr 27min) was the longest and most personal Altman interview on record — conducted four months after his firing and reinstatement and covering the events of November 2023 in detail for the first time. Altman's account of the board crisis: he was surprised by the firing, believes the board's concerns were not grounded in documented evidence of wrongdoing, and thinks the episode revealed structural weaknesses in how nonprofit boards govern dual-use technology companies. He declines to specify what the board's stated concerns actually were, but says "they were wrong." The conversation's most substantive sections are on o1 development (which was already underway), the competitive landscape after GPT-4, and Altman's updated views on AGI timelines — he is notably more specific in 2024 than he was in 2023, suggesting he believes capabilities are advancing faster than he expected. His framing of the relationship between Anthropic and OpenAI shifted: in 2023 he described them as colleagues with different safety philosophies; in 2024 his language was more competitive. On the governance question, Altman announced changes to the board structure and committed to hiring a more diverse board with relevant operational experience. This interview is the primary source for understanding Altman's subjective account of the most important governance crisis in AI history.