Faggella argues scoffing at AGI risk is no longer intellectually honest, citing Bengio and Hinton
danfaggella · x · 2026-10-03
In a spat with Gary Marcus and epidemiologist Angie Rasmussen, Daniel Faggella shares his essay arguing that outright dismissing AGI and its risks is no longer an intellectually defensible position in 2025.
- He points to shifting expert consensus: deep learning pioneers Bengio and Hinton openly warn of near-term AGI risk, and Yann LeCun — the most credible skeptic of rapid progress — now says human-level AI may be only 5-10 years away.
- Virtually all major lab leaders (Altman, Hassabis, and others) have signed the statement on AI risk.
- He criticizes treating "AGI is impossible" as fact rather than hypothesis, noting this rebuttal might have worked in 2015 but no longer fits the landscape.
Gary Marcus replied that he is not writing off AGI risk entirely, that he believes current AI poses many risks, and that his original critique targeted Hinton's inconsistency rather than Bengio, whom he considers cogent.
Related event: Marcus and Faggella Debate AGI Risks as Bengio Warns UN of Unstoppable AI(3 posts)→
More from AGI Musings
- LeCun Says He Has 'Zero Concerns' About Human Extinction, Calls Amodei 'Deluded' — marigo · 2026-10-03
- Hamming's classic 'You and Your Research': why so few scientists matter — joshua_saxe · 2026-10-03
- AI consciousness will enter the Overton window within years, says Joshua Saxe — joshua_saxe · 2026-10-03
- Stephen Wolfram on the future of pure math research in the age of AI — pickover · 2026-10-03
- "Taste is genuinely one of the final moats" in the age of commoditized AI — eptwts · 2026-10-03
- Gary Marcus mocks LLM consciousness claims: 'math equations are not conscious' — GaryMarcus · 2026-10-03