descriptive claim
Aschenbrenner forecasts that AGI by 2027 is strikingly plausible, projecting another preschooler-to-high-schooler-sized qualitative jump from GPT-4 by 2027 via ~0.5 OOMs/year in compute, ~0.5 OOMs/year in algorithmic efficiency, and 'unhobbling' gains (chatbot to agent).
desc_aschenbrenner_agi_2027_forecast
confidence 0.90
Evidence (1)
supports (1)
- SITUATIONAL AWARENESS: The Decade Ahead --- Introduction expert_estimateweight0.90
locator: Table of Contents, Section I summary
“AGI by 2027 is strikingly plausible. GPT-2 to GPT-4 took us from ~preschooler to ~smart high-schooler abilities in 4 years.”