Has Theoretically-Guided Practice Vanished in Modern Machine Learning?
NeighborhoodFatCat · reddit · 2026-08-15
A Reddit discussion argues that many classic ML theories (e.g., overfitting risks, generalization limits for big models, test set taboos) have been debunked or rendered irrelevant by practice. These theories evolved from math to folklore in textbooks and interviews, only to be overturned by the empirical success of deep learning. The post questions if ML is now purely empirical or if any theoretically-guided practices remain.
More from Research
- Epoch researcher lists top questions governing AI's global impact — Jsevillamol · 2026-08-15
- Yukon accelerates open research with intelligent agents — gajesh · 2026-08-15
- TEMPO: Solving Long-Horizon Agent Training via Macro-Step Policy Optimization — teortaxesTex · 2026-08-15
- Code World Model moves beyond static code prediction with internal world models — bendee983 · 2026-08-15
- Academic Discussion on the Term "Mechanistic World Model" — sirbayes · 2026-08-15
- Study: Gemini, ChatGPT, and Claude Hallucinate on Missing Information — Plastic-Cell-4497 · 2026-08-15