Gary Marcus clarifies: he's not dismissing AGI risk, only Hinton's inconsistent claims
GaryMarcus · x · 2026-10-03
Responding to Daniel Faggella's essay arguing that scoffing at AGI is no longer intellectually honest, Gary Marcus clarifies his position on two points.
- His original tweet targeted Hinton, not Bengio: he calls Bengio cogent while describing Hinton as "all over the map."
- He denies writing off AGI risk entirely, stating he has repeatedly said current AI poses many risks.
The exchange highlights a nuance in the AGI debate: Marcus's skepticism is about the consistency of specific prominent voices rather than a denial that AI risk exists.
Related event: Marcus and Faggella Debate AGI Risks as Bengio Warns UN of Unstoppable AI(3 posts)→
More from AGI Musings
- Stop selling AI engineering software, start building manufacturing capacity, says founder — IgorCarron · 2026-10-03
- Google's Jon Barron: models should never matter morally, today and forever — tobias_rees · 2026-10-03
- Google's James Manyika: The AI Industry Can't Police Itself Alone — bloomberg · 2026-10-03
- PhD with 12 ML papers, 5 at top conferences, asks if research roles are out of reach — Hope999991 · 2026-10-03
- David Krueger's talk blasts "managing the transition to post-AGI" framing via Martian takeover analogy — conitzer · 2026-10-03
- "Agent John Rawls": designing agent society you'd accept being the median session of — zetalyrae · 2026-10-03