Both Sides Are Overconfident On AI | Toby Ord

Ord estimates a power-law exponent (λ≈0.5–0.6) governing agent-swarm scaling that could determine whether recursive self-improvement explodes or plateaus.

• Swarm scaling law: Ord derives a power-law relating agent-swarm size to effective compute, with exponent lambda (λ) around 0.5–0.6 estimated from OpenAI data and Opus 5.5's system card, where some tasks (e.g., independent documentation writing) hit λ≈1 (near-perfect parallelizability).
• RSI implications: This λ maps directly to economists' growth-theory parameter R; R=1 yields exponential growth, R>1 implies runaway growth reaching infinity in finite time absent real-world constraints.

Themes, sentiment and reasons are AI-written paraphrases, checked so they never quote an episode. They may contain errors. Not investment advice. Episodes remain the property of their creators.