Gary Marcus: Expected-Value Math Shows Anthropic Can't Justify a $2T Valuation
GaryMarcus · x · 2026-09-30
Gary Marcus argues Anthropic-style AI valuations ignore basic expected-utility math: promising to own a $30 trillion market at 0.01% odds yields just $30 billion in expected value — nowhere near a $2 trillion valuation. He says the math only works by ignoring failure modes: massive liability, competition, price wars, rising rates, open source, and the eventual shift to local models. Applying the same logic to existential risk, he notes a 10% chance of annihilating humanity means an expected loss of 800 million lives.
More from AGI Musings
- Why Mythology, Not Sci-Fi, May Be the Better Lens for Understanding AI — begusgasper · 2026-09-30
- Nature comment: AI boosts individual productivity but narrows scientific exploration — begusgasper · 2026-09-30
- Study: 'fit' interviews elicit social-position signals via family and upbringing topics — soumitrashukla9 · 2026-09-30
- Software Is Dead, Long Live Data-for-Synthesis: Reddit Debates the Agent Era — Happy_Conference8401 · 2026-09-30
- AI Helped Crack a 1967 Fusion Plasma Conjecture, Finding an Exact 3D Asymmetric Solution — Dr_Singularity · 2026-09-30
- The Underappreciated AGI Joke: A Baby That Could Talk Would Be Worth Many Hogs — S_OhEigeartaigh · 2026-09-30