Redis vs Keras Authors Debate: Deep Inference or Mechanical Parroting in LLMs?
antirez · x · 2026-08-03
Redis author antirez continues the debate with Keras author François Chollet regarding the capability boundaries of Large Language Models (LLMs).
antirez clarifies that the skeptics' argument is nuanced: even within the local space represented by the training set, LLMs cannot interpolate based on a deeply inferred representation. Instead, they only interpolate on a shallow, almost-already-seen realm, which aligns with the "parrot model" critique.
More from AGI Musings
- Musk Predicts AI Will Exceed Total Human Intelligence Within 5 Years — Saboo_Shubham_ · 2026-08-03
- Anthropic's AI Fable Proves Erdos #146 False, Mathematicians Find No Obvious Errors — ctjlewis · 2026-08-03
- Eric Horvitz & Robert West Warn the Window for Aligned, Accountable AI is Narrowing — erichorvitz · 2026-08-03
- Why the Public Laughs Off AI Existential Risk — AndrewCritchPhD · 2026-08-03
- AI Solving Math Isn't a Tragedy: The Goal of Science is Truth, Not Jobs — garrytan · 2026-08-03
- Credit in AI-Assisted Discovery: Humans or Tools? — Leather_Area_2301 · 2026-08-03