“I would argue that Coxon is exaggerating what AI is likely to do anytime soon”
Marcus disputes Coxon's near-term capability timeline and considers extinction unlikely, but finds his firsthand account of frontier-lab thinking plausible and troubling. He says catastrophic harm remains possible and rejects the idea that merely developing the technology first solves the danger.
- Speaker
- Gary Marcus
- Capacity
- AI researcher and author; signed personal commentary in Marcus on AI.
- Observed
- September 10, 2026
- Published here
- September 10, 2026
- Classification rationale
- Disagrees over timeline and extinction likelihood while supporting criticism of lab reasoning.
Surrounding material
Full context
Before the quotation
Part II explicitly discusses Coxon's resignation thread. Marcus says it emphasizes future harms too heavily relative to current harms and distinguishes it from the separate Terence Tao controversy covered earlier in the article.
After the quotation
Marcus discusses Hubinger's response, criticizes the lack of a serious plan, and argues that a different technical foundation might help. These are Marcus's assessments, not DI findings.
Evidence
Complete source record
Full primary article; Part II discusses Coxon. Publication date is displayed without an exact time. Speaker and position scope are explained in the linked event.