Scott Aaronson (Shtetl-Optimized) (@scottaaronson)
Schlumberger Centennial Chair of Computer Science at UT Austin. Co-author (with Boaz Barak) of 'Five Worlds of AI' (April 2023), adapted from Russell Impagliazzo's 1995 Five Worlds of Complexity.
Shtetl-Optimized • UT Austin CS • OpenAI Safety Alum
Immutable Forecast Track Record (0)
Authored & Cited Arguments (3)
Even if scaling hit a wall tomorrow, integrating existing frontier models into the global economy already virtually rules out a true 'AI-Fizzle' by 2050.
When Boaz Barak and I defined AI-Fizzle in April 2023, we compared it to nuclear power or supersonic flight—technologies that worked but plateaued below a general societal revolution. With 24+ years remaining until 2050 and agentic coding/research systems already transforming software and science in 2026, the burden of proof for a 25-year global stall is extraordinarily high.
Singularia (17%) represents the world where superintelligence is achieved and acts like 'benevolent gods'—solving disease, aging, and climate change beyond human comprehension.
In our taxonomy, Singularia is distinguished from Futurama by whether baseline human history recognizably continues. If an intelligence explosion happens rapidly enough that the annual panel never settles on Futurama first, and our alignment guardrails hold, humanity enters Singularia directly.
AI-powered surveillance states remove the historical check on tyranny: the need for human enforcers with consciences.
In 'Five Worlds of AI', we highlighted how an authoritarian regime armed with ubiquitous AI vision, voice analysis, and loyal robot enforcement could achieve a permanently stable Orwellian panopticon.