Smaller experiments win: the Kaggle lesson for doing more AI research with less compute
strickvl · x · 2026-11-21
Will Brown argues you can do far more research with any compute budget if you just make experiments smaller. A replier connects it to the pre-LLM Kaggle era: winners ran the most experiments, mastering how to pick the smallest model and data size that still gave a noisy but directionally correct, statistically trustable Bayesian update. The core lesson: information gain per unit time is the name of the game.
More from AGI Musings
- Terence Tao: AI is winning at math but we've learned almost nothing mathematically — GaryMarcus · 2026-11-21
- Ramez Naam extrapolation: AI a decade away from completing 98% of hour-long human tasks — GaryMarcus · 2026-11-21
- Shermer doubles down: AI means protopia, not utopia or dystopia — intellectronica · 2026-11-21
- Hassabis predicted 2 years for AI to crack a Millennium Prize problem; 6 months later, both Riemann and Navier–Stokes have proposed solutions — Hesamation · 2026-11-21
- Terence Tao on AI math: 'strip mining' that wins proofs but teaches us nothing — yacineMTB · 2026-11-21
- Imagine P=NP: Verifiable Tasks Solved by Compute, Security in Tatters — gleech · 2026-11-21