Scott Alexander: 'The Sigmoids Won't Save You' — AI Scaling Arguments Dissected — 282 Points on HN
Astral Codex Ten / Hacker News·medium signal
Scott Alexander's new Astral Codex Ten essay tackles the common AI skeptic argument that 'all exponentials eventually become sigmoids' (S-curves that flatten). He argues that while technically true, this doesn't mean flattening happens at the moment you're analyzing. The key insight: predicting WHEN growth flattens requires fully understanding the underlying process, not pattern-matching to historical S-curves. The post generated 266 comments with practitioners debating current AI scaling trajectories.