Author self-corrects: saddle point analogy explains fitting, not generalization
ianand · x · 2026-11-17
ianand self-corrects his earlier analogy: technically it explains why deep learning can be trained to fit data, not why it generalizes — a crucial difference. We test AI models' generalization with held-out data; your relative doesn't.
Related event: Saddle Point Analogy Explains Why Conspiracy Theories Keep Adapting(2 posts)→