Order the jagged AI frontier by data volume and it becomes a smooth spiral, argues ex-Google XC lead
rickasaurus · x · 2026-10-15
awiltschko offers a reframing of AI's "jagged frontier": model capability tracks data availability, not some smooth general progress.
- The frontier's spikes are tasks with massive data (math, programming, games — static or simulatable); the valleys are data-scarce tasks (cooking lasagna, olfaction, manufacturing a passenger train). He knows of no exceptions
- Reorder tasks by data volume and the jagged frontier becomes a smooth spiral: superintelligent where data is abundant, progressively weaker as it thins
- ramez Naam adds that AI capability depends on data volume, accuracy, and the ability to do perfect validation in compute (e.g., RL environments), with a four-criteria definition of "highly verifiable": error-free evaluation, unlimited data volume, machine-speed feedback, and measuring what we actually care about
- Accompanying charts map tasks from abundant static data through closed-loop simulators to scarce/expensive data
Related event: AI's Jagged Capability Frontier Reframed as a Function of Data Volume(2 posts)→
More from AGI Musings
- Ex-OpenAI interpretability lead: most neurons in trillion-parameter models are now legible — burny_tech · 2026-10-16
- MIT's Buehler on AI as a new language of science: superintelligent partners and executable papers — CatAstro_Piyush · 2026-10-16
- SEO visibility now tracks impressions, not clicks, as AI Overviews eat traffic — lilyraynyc · 2026-10-16
- Google white paper: AI river-flood forecasts 7 days out for 2B+ people, FireSat wildfire tracking — ymatias · 2026-10-16
- Polymarket odds: Only 10% chance US enacts AI safety bill by end of 2026 — Polymarket · 2026-10-16
- Intelligence is proportional to data: reorder tasks by data volume and the jagged frontier becomes a spiral — sebkrier · 2026-10-16