antirez: Calling classifiers 'decision models' is a crime against machine learning
antirez · x · 2026-10-03
SQLite creator antirez pushed back on calling classifiers "decision models," arguing it's a misreading of machine learning. His core point: these models decide far less than LLMs, since chain-of-thought can be seen as serial information processing that classifiers lack. Such models are either small LLMs that aren't allowed to think (using logits at the last prefill position to categorize) or BERT-like models that compress input meaning and project a class — in both cases deciding much less than an LLM.
More from Models
- Physician: AI competition now measured in weeks, GPT-6 versions 7 days apart — ramos_casals · 2026-11-19
- KradleBench pits humans and AI in live multi-agent Minecraft evals — lmoroney · 2026-11-19
- micro1's flow qc model screens training data for benchmark leaks before release — Exp_Mark · 2026-11-19
- Gemini 4 Neon appears: frontier intelligence at $3/$15 per M tokens — Angaisb_ · 2026-11-19
- Leaker predicts new DeepSeek paper this week, scheduled for November 11 — teortaxesTex · 2026-11-19
- Gemini 4 Neon rollout begins: leaked UI shows 3-tier Gemini 4 lineup with Carbon locked — testingcatalog · 2026-11-19