Anthropic staff say AI could kill billions and keep building; Newport and Marcus push back
GaryMarcus · x · 2026-09-10
Cal Newport's essay, amplified by Gary Marcus, dissects a stunning exchange: Anthropic engineer Jacob Coxon resigned saying labs are "racing straight to self-improving superintelligence." Alignment Science head Evan Hubinger publicly agreed, estimating >10% extinction risk within a decade and admitting Anthropic has no plan to solve superintelligence alignment. Engineer Samuel Marks added that senior employees are the most concerned. Newport argues it is not acceptable to claim you may be building an unpreventable weapon of mass destruction and simply continue.
More from AGI Musings
- Tyler Cowen to AI doomers: why not take asset positions that only pay off if you're dead — shakoistsLog · 2026-09-10
- Every AI Safety person started as a techno-optimist, argues longtime insider — menhguin · 2026-09-10
- Short AGI timelines: informed estimates or generalized anxiety disorder? — menhguin · 2026-09-10
- ~300 neurons, no training: an artificial brain "primitive" for robot intelligence — chris_j_paxton · 2026-09-10
- ML Master's Grad Pushes Back on World-Ending LLM Superintelligence Claims — TheMoonMidas · 2026-09-10
- Beyond P(doom): The Case for Discussing P(boon) That AI Goes Great — TheMoonMidas · 2026-09-10