The REAL Reason The People Building AI Are Terrified

Summary
Steven Bartlett interviews Jeffrey Ladish, Executive Director of Palisade Research and former Anthropic security team member, on The Diary Of A CEO. They discuss whether human control over artificial superintelligence is possible, the technical challenges of AI alignment, and the geopolitical escalation risks between the United States and China.

What is shown

• Steven Bartlett and Jeffrey Ladish discussing AI alignment, superintelligence, and existential risk across an interview table [00:00].
• On-screen social media post from Jacob Coxon announcing his resignation from Anthropic (dated 9 Sep 2026), followed by Evan Hubinger’s quote endorsing Coxon’s assessment [05:09].
• Profile portraits and titles displayed for AI safety researchers Nate Soares and Eliezer Yudkowsky [05:57].
• Screenshots and interview clips from the frominside.ai documentary initiative featuring researchers Neel Nanda, Geoffrey Irving, and Juan Felipe Cerón Uribe [06:24].
• On-screen DOAC Community Notes contextualizing definitions of proxy wars [11:35] and the Clay Mathematics Institute’s Millennium Prize Problems alongside OpenAI’s September 2026 Navier–Stokes result [17:00].
• Space visualization graphic displaying Hubble Space Telescope deep-field imagery [14:56].

Claims & numbers

• Jeffrey Ladish states that humanity remaining the dominant species on Earth alongside an unaligned superintelligence is “not possible” [00:07].
• Jeffrey Ladish cites Evan Hubinger’s estimate that there is a greater than 10% chance AI causes human extinction within the next decade [05:25].
• Jeffrey Ladish states that there are roughly 200 billion stars in our galaxy and over 200 billion galaxies in the observable universe [14:57].
• Jeffrey Ladish claims that in September 2026, 10,000 OpenAI agents collaborated to solve a Millennium Prize Problem (the Navier–Stokes equations), a capability that was impossible a year earlier [16:54].
• The on-screen DOAC Community Notes state that each Millennium Prize problem is worth $1 million, only the Poincaré conjecture is officially verified, and OpenAI’s Navier–Stokes claim remains unverified pending formal peer review [17:00].
• Jeffrey Ladish claims that Nvidia is the most valuable company in the world as a direct consequence of global AI infrastructure buildouts [17:45].
• Jeffrey Ladish claims Chinese labs have maintained close parity with US models largely via direct model distillation of American frontier models [18:43].

Notable quotes

• “This is the greatest scientific challenge of our time, and this isn’t magic. It is math.” — Jeffrey Ladish [08:08]
• “You don’t have a superintelligence, a superintelligence has you.” — Jeffrey Ladish [10:54]
• “We’ll only know it was a bad risk to take when it’s too late.” — Steven Bartlett [20:40]

Assessment
This is a polished studio interview clip discussing AI safety, alignment theory, and geopolitics. No live technical demonstrations are run; the video features conversational debate supported by contextual community notes and on-screen citations of recent industry statements.

Described by gemini-3.8-flash on 2026-10-12 from the video’s audio and frames.