OpenAI researcher invokes 2019 double descent paper in scaling-law spat
willdepue · x · 2026-11-06
OpenAI's willdepue posts a screenshot of the 2019 Harvard double descent paper's experimental setup to counter a claim about the "data wall", mocking that the cited evidence rests on tiny 5-layer ResNets.
The thread hinges on a commenter's framing: "either it is real or OAI isn't real" — the interpolation-regime theory and OpenAI's scaling approach stand or fall together. The screenshot shows the paper's experiments used width-scaled ResNet18s, 5-layer CNNs and 6-layer Transformers, all tiny by today's standards.
Related event: OpenAI Researcher Disputes Double Descent Argument in Data Wall Debate(4 posts)→
More from AGI Musings
- Don't defer to Twitter's sub-hour takes on METR — read the source — gleech · 2026-11-06
- EA moral views debate: critics accused of relying on even more extreme moral premises — ShakeelHashim · 2026-11-06
- Veteran dev: AI-built TraceDraw beats CorelDRAW 2026, predicts software industry upheaval by 2036 — Maxiol777 · 2026-11-06
- AI self-help clickbait everywhere sparks Fahrenheit 451 comparison: Hamlet as a one-page summary — chalmermagne · 2026-11-06
- X debate: does an LLM's 'thinking' have agency outside a prompt-triggered loop? — teodorio · 2026-11-06
- AI circle debate: can artists who shun AI still be called great artists? — repligate · 2026-11-06