AI 2027 authors spooked: Anthropic may imminently hit RSP 2x acceleration, ASI sooner than predicted
shadowt1tan · reddit · 2026-11-04
Daniel Kokotajlo (AIFP, co-author of AI 2027) responds to Theo Bearman's finding that Anthropic updated its S1 with notable additions on automated AI R&D and human oversight, suggesting Anthropic may be 'imminent' to cross its RSP threshold of 2x acceleration over baseline.
- Kokotajlo says AIFP is 'spooked': if Anthropic really is about to hit that threshold, timelines to full RSI and superintelligence may be very short
- Naively optimizing the AIFM parameters for 2x acceleration in Jan 2027 yields superintelligence sooner than depicted in AI 2027
- The team promises a lengthier update on their thinking within a week
Related event: Anthropic Updates RSP, May Be Near AI R&D Acceleration Threshold(3 posts)→
More from AGI Musings
- Time to Post: why lower latency from thought to expression feels more authentic — signulll · 2026-11-04
- Terence Tao warns of real-world damage from delegating thinking to AI soon — rohanpaul_ai · 2026-11-04
- 'My agent deleted it, but I don't believe that': an LLM take on Moore's paradox — jessyjli · 2026-11-04
- Embryo Screening for Depression Genes Sparks Backlash, Called Pseudoscience — nwilliams030 · 2026-11-04
- Eric Jang: Blender and Lean emerged as this year's dark-horse tools in AI — ericjang11 · 2026-11-04
- Your Brain Predicts Too: A Reddit Case That Humans and AI Are More Alike Than We Admit — Ok_Translator_227 · 2026-11-04