AI 2027 authors now expect a Superhuman AI Researcher in 2027, citing METR’s finding of ~2x faster AI R&D at Anthropic
Confirmed
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- 3 of 5
- Last verified
- 12 November 2026
Your AI and this story
- GPT-6.2 Astra163 days after its cutoff
- Claude Opus 5.5133 days after its cutoff
- Gemini 4 Hydrogen224 days after its cutoff
- Grok 4.7163 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 133 days before it.
Key facts
- Forecast: SAR ‘will likely arrive at some point in 2027’; 35-65% chance of Top-Expert-Dominating AI (TED-AI) before 2028; Eli Lifland’s medians: SAR August 2027, TED-AI September 2028
- Before: Kokotajlo had put 50% on SAR before 2029; the previous update was less than three months earlier
- Why: METR’s review of Anthropic’s risk report put AI R&D progress at about 2x the historical rate, ~80% attributed to AI; the authors read this as ~1.7x AI-attributable acceleration, about 5 months ahead of the AI 2027 scenario, which predicted 2x uplift in December 2026; Kokotajlo’s model had given under 4% to this much acceleration
- Caveat they state: much of METR’s evidence is redacted, so the update depends on trusting unseen analysis
- Key line: ‘If SAR is coming in 2027, it is incredibly urgent to pause or substantially pace frontier AI development’
- Reactions (Zvi Mowshowitz, Nov 9): Ryan Greenblatt said he was ‘seriously wrong’ (he expected 1.15-1.3x for 2026); Toby Ord estimates ~1.4x if a one-time jump from the Mythos pretraining run is counted; Anthropic’s Sam McCandlish said Anthropic will publish its own analysis in the coming weeks; Rep. Alex Bores called for banning recursive self-improvement
What happened
The update is the group’s response to METR’s Nov 5 review of Anthropic’s automated-R&D risk report. It sets out each author’s view and an illustrative trajectory; it is a forecast, not a measurement.
Why it matters
The AI 2027 scenario was the most widely read forecast of an intelligence explosion. Its authors had pushed their dates later during 2025-26; this update reverses that on the strength of a third-party measurement of a lab’s own R&D speed-up.
Sources
3 sources from 3 sites. Numbers match the chips in the text.
3 sources: 1 primary, 1 press, 1 reaction
Primary
- AI Futures Project: Q3.5 timelines updateblog.aifutures.org, official
Press
- Crypto Briefing: AI 2027 authors now expect a superhuman AI researcher in 2027cryptobriefing.com, press
Reactions
- Zvi Mowshowitz: AI R&D is accelerating towards potential RSI in 2027thezvi.substack.com, discussion
Changes
- Filed (news agent)