Rewind: Ronan Farrow on Sam Altman’s ‘unconstrained’ relationship with the truth

The episode begins with Nilay discussing the episode's revisit of a conversation from earlier in the year with Ronan Farrow, an investigative journalist and New Yorker contributor. The discussion revolves around Farrow's detailed piece co-authored with Andrew Marantz, which profiles Sam Altman, CEO of OpenAI, and explores his controversial reputation for deception and the broader issues in AI industry accountability. Nilay emphasizes the relevance of Farrow's reporting as public concern about unregulated rogue AI incidents increases and highlights the importance of its insights for current debates on AI oversight.

Farrow explains that it took a year and a half of meticulous investigation, talking to over a hundred sources, to assemble their deep profile of Sam Altman. He notes that Altman's reputation has shifted over time, from being viewed as a promising leader to a figure criticized for dishonesty and sociopathic traits, according to some insiders. Farrow highlights how Altman's behavior and the culture at OpenAI reflect a broader race to the bottom in safety standards, with safety concerns often being obscured or downplayed, especially amid little political appetite for regulation.

Farrow presents findings from the investigation showing that OpenAI's internal reports and legal investigations, including one by Wilmer Hale, were kept out of public and internal documentation, with crucial details only presented orally. Company insiders describe Altman as a pathological liar and sociopath, with some telling Farrow that his traits of people-pleasing and deception enable him to rally conflicting constituencies but also cause dysfunction. Farrow discusses how Altman's self-assessment reveals a reluctance to confront his flaws, which contributes to ongoing safety and trust issues, and how Silicon Valley hype culture fosters and enables such traits.

Farrow explores how the race for AI dominance and profitability affects industry behavior and public policy. He notes that despite the dangers, regulatory agencies remain weak and company influence over politicians is profound. Big tech's extensive lobbying, including contributions from executives like Greg Brockman, hampers meaningful oversight. Farrow advocates for external regulation, increased safety testing, whistleblower protections, and calls for voters to scrutinize political donations from AI industry players, emphasizing the need for political accountability to counteract the industry's influence.

Farrow reflects on whether OpenAI and other AI companies will reach their goal of an IPO, questioning if they can 'flip the switch' to profitability amidst internal doubts about leadership. He discusses a stark comparison made by a Microsoft executive, suggesting Altman's legacy could resemble that of Bernie Madoff or Sam Bankman-Fried if his deception leads to a downfall. Farrow discusses how market pressures and ideological risks, such as unchecked power and lack of accountability, threaten the industry's future. He emphasizes that the current bubble and investment dynamics could burst, halting AI's rapid growth and forcing companies to confront ethical and safety issues.

In closing, Farrow and Nilay revisit the essential question of industry accountability and the difficulty of managing the powerful incentives at play. Farrow highlights that Silicon Valley elites often believe they can control the future without external oversight, driven by a nihilistic attitude that dismisses the social contract. He advocates for more robust regulation and legal protections, like whistleblower laws, to combat corporate secrecy. Farrow underscores that public opinion increasingly fears AI risks, but political and industry interests remain deeply intertwined, making external regulation vital. The episode ends with cautious optimism about the potential for collective action and voting to shape industry accountability.