Report: OpenAI has largely automated training of new experimental AI models
zephyr_z9 · x · 2026-09-21
- OpenAI's internal models now handle much of building and training experimental models, including writing GPU kernels and optimizing run code
- Researchers give a single optimization example, then let AI work for weeks implementing and testing similar improvements
- Internal agents increasingly collaborate with each other without human involvement; capability improved sharply in recent months, aided by growing compute access
Related event: Report: OpenAI Has Automated Most Experimental Model Training(3 posts)→
More from AGI Musings
- Toby Ord on swarm scaling: 10,000-agent run cost ~$20M, solved Navier-Stokes in 88 hours — tobyordoxford · 2026-09-22
- Krueger: "indefinite" AI moratorium doesn't mean permanent — DavidSKrueger · 2026-09-22
- Toby Ord: Agent Swarm Estimates Put Intelligence Explosion Parameter λ at 0.5-0.6 — tobyordoxford · 2026-09-22
- Why did Altman, Amodei and Musk suddenly agree on pacing frontier AI within two weeks? — k1_r1 · 2026-09-22
- Agent swarm data suggests intelligence explosion parameter λ around 0.5-0.6, below threshold — tobyordoxford · 2026-09-22
- Amazon blocks Muse: platforms brace for an agent-mediated power struggle — tobowers · 2026-09-22