AI’s apparent generality exceeds what headline metrics justify

Evidence

Mitchell argues that contemporary AI’s apparent generality is more limited than headline metrics suggest, and that core questions about understanding and meaning remain unresolved. — , The Braided Reading EPUB, 1783

Categorization

• XREF: Corroborates Marcus and Davis’s critique of deep learning generalization limits in [1787]. — connects to: Training models on absurd datasets exposes AI’s true capabilities — , The Braided Reading EPUB: 5-Year…, 1863; Hype can mislead public perception of AI technology. — Emily M. Bender, Alex Hanna, The AI Con- How to Fight Big Tech’s…, 10; Context-predicting models obtain a thorough and resounding victory against count-based counterparts — Glove: Global Vectors for Word Representation (+4 more)

Broader Connections

• Hype can mislead public perception of AI technology. #corroborates (0.60) — Emily M. Bender, Alex Hanna, The AI Con- How to Fight Big Tech’s…, 10
• Even randomly weighted linear models beat human expert judgments #exemplifies (0.55) — Brian Christian, The Alignment Problem, 441