FULL STORY

AI Doom Warnings Spark e/acc Counterattack

After a DeepMind researcher warned of AI extinction risk, the industry erupted in debate over risk narratives. e/acc founder Beff Jezos fired back, accusing the AI Safety camp of enabling regulatory capture by big tech.

2026-09-09 ~ 2026-09-14 · 4 episodes · 31 posts

Episode 1 · Debate Rages Over AI Doom Narratives and Control Agendas (2026-09-09, 11 posts)

A multi-author clash erupted on the timeline over whether AI doomerism is valid and what motivates its advocates. No consensus emerged, but the debate shifted from "do risks exist" to "who gets to define risk and for what policy purposes."

Confirmed

  • markk posted a long critique accusing AI doomers of exploiting public discontent over datacenter expansion, arguing their real goal is controlling who can build and use AI and how fast progress goes; he cited Schachtel's view that any AI-slowing agreement must account for competitors like China. LeCun amplified Dan Jeffries' argument that doomers' remedies amount to rigid centralized control or progress-stalling economic decline.
  • markk also declared his accelerationist stance: tired of the "psychological warfare," he said staying silent risks losing the freedom to use AI to doomer-driven limits and bans. He later joked that dying at the hands of an angry doomer mob is likelier than dying from AI.
  • Dan Jeffries cited Andrew Yang's failed job-pocalypse predictions, arguing fear itself is the goal—a tool for control and votes—and that a technology used safely by billions deserves evidence-based scrutiny, not appeals to crowd emotion. Jim Pethokoukis mocked experts who predict extinction fluently but fail at predicting unemployment.
  • SydSteyerhart claimed "AI safety" rhetoric is really about power: elites fear technological displacement and the collapse of bureaucratic control.
  • alexolegimas countered that those closest to the technology have the best forecasting record on capabilities; AgustinLebron3 split the difference—capability forecasts are accurate, economic-impact forecasts are not.
  • Researcher apsarathchandar offered a middle path: he studies real current risks (bias, fairness, misinformation) but does not believe AI will kill everyone, calling extinction-rescue claims self-important.

Why it matters

  • The core dispute is about who defines risk and how risk narratives serve policy agendas—one side fears centralized control and stagnation via regulation, the other cites insiders' track record of warnings.
  • The datacenter backlash is the concrete battleground; markk says the stakes are too high to stay silent.
  • Amplification by figures like LeCun shows the debate has entered mainstream AI discourse, with forecasting records (extinction vs. unemployment) becoming the key credibility weapon on both sides.
  • apsarathchandar's stance shows a "real but non-extinction" middle path is gaining space between the two poles.

Episode 2 · DeepMind Researcher Publicly Warns of AI Existential Risk, Sparking Debate (2026-09-10, 4 posts)

A Google DeepMind researcher publicly voiced fears that AI could pose existential risks, urging alignment research to far exceed capability research, while other researchers countered that scaling current LLMs won't cause qualitative change and that the real danger lies in malicious humans misusing AI rather than the technology itself.

Episode 3 · e/acc Founder Beff Jezos Launches All-Out Counterattack on AI Doomers (2026-09-12, 14 posts)

In mid-September 2023, e/acc (effective accelerationism) founder Beff Jezos (Extropic CEO Guillaume Verdon) posted intensively on X from 09-12 to 09-13, launching a full counterattack on the "AI Doomer" camp. His central thesis: spreading panic is itself the real danger, potentially greater than AI itself.

Confirmed

  • Core statements: panic is the real danger and spreading fear is not a virtue (m2); doomers are "injecting cortisol into everyone, spreading their anxiety" (m8); sharing David Friedberg's interview, he called "decel psyop panic" the true threat to social welfare, noting panic-makers never pay for long-term damage and missed opportunities (m10).
  • On 09-13 he posted a long piece alleging the week's anti-AI events were a coordinated PR operation, citing a timeline: a doomer post algorithmically boosted to 120M+ views, Democratic politicians amplifying AI fear the same day, mainstream media coverage, MIRI's Nate Soares on Tucker, UK regulators' joint letter, etc. (m1).
  • Citing Martin Casado, he called a certain AI-safety-related release "the doomers' 9/11," implying premeditation (m7). He also endorsed Suzanne Scanziani's satire mocking doomers for assuming superintelligence can magically solve all physical constraints, with overnight foom violating physics (m9).
  • Amplifying Ramez's open-source AI summit demo, he summarized: closed-source labs are coordinating mass media campaigns and lobbying politicians to stop open-source AI—"they don't want to compete in the free market"—a "manufacturing consent, centralizing power" movement (m11).
  • He stated p(1984) far exceeds p(AI Doom), expressing alarm at the regulatory climate (m13).
  • He targeted the AI safety ecosystem: "AI Safety is the new woke," accusing an AI-doomer NGO industrial complex of becoming a new DEI-style NGO industry (m6). Citing Kevin Bass, he said "AI safety means the opposite of what its advocates claim," mocking those who sell p(Doom) while guaranteeing p(1984)=1 (m4).
  • In cruder posts, he mocked doomers as low-testosterone, weak, dependent on tribal authority and naturally anxious, joking that "testosterone in the water" would solve the AI doomer problem (m3).
  • Sympathizers: MickeySteamboat cited Beff's "Doomer is the new woke" to accuse doomers of coordinating, even illegally, to slow AI (m14); a long anti-doomer essay criticized self-appointed AI guardians for packaging research delays as "prudence," power concentration as "safety," censorship as "risk reduction," while ignoring the measurable annual costs of stagnation (m5); another thread accused parts of the MIRI school of advocating authoritarian world government by a council of "super-rationalists" (m12).

Unconfirmed

  • Claims of a "coordinated PR operation" and "closed-source labs coordinating media and lobbying" remain one-sided e/acc narratives based on timeline coincidences and motive inference, without direct evidence of organized coordination.
  • The description of MIRI advocating authoritarian world government comes from secondhand commentary and may not accurately represent MIRI's positions.

Why it matters

This was a landmark exchange in the e/acc vs. AI-safety debate: one side invokes existential risk to justify slowdown and regulation; the other fires back citing stagnation costs, power concentration, and narrative manipulation. As the flagbearer of e/acc and CEO of Extropic, Beff's "p(1984) > p(doom)" framing and "AI Safety is the new woke" charge gave the anti-regulation camp punchy slogans, directly shaping the subsequent public debate over open vs. closed source and acceleration vs. safety.

Episode 4 · Beff Jezos Slams AI Safety Camp as Regulatory Capture Tool (2026-09-14, 2 posts)

e/acc founder Guillaume Verdon (Beff Jezos) accused the AI Safety camp of running a psyop and enabling regulatory capture, arguing that doom warnings make safety advocates 'useful idiots' helping incumbents dominate the AI market.