Burkov: Recursive AI Self-Improvement Will Look Funny at 3 Months per Training Run
burkov · x · 2026-10-03
Andriy Burkov, author of The Hundred-Page Machine Learning Book, quips that recursive AI self-improvement will look very funny in practice, since training each recursively improved model version takes about 3 months — a jab at how slow training cycles would bottleneck any self-improvement loop.
More from AGI Musings
- A 5-Year-Old in Shenzhen Is Building a POS System for His Dad's Dumpling Shop — alexmacgregor__ · 2026-10-03
- Personal Agent Demos Are Stuck on Restaurant Bookings While Users Want Taxes and Invoices Handled — sujingshen · 2026-10-03
- Benjamin Bratton on what humans have that AI doesn't: babies, dreams, anesthesia, crowds — bratton · 2026-10-03
- 'One thought too many': where should formula-based AI decisions stop? — rasta321 · 2026-10-03
- Burkov Mocks Open-Weight Security Panic: Hackers Could Spawn 'Millions of Viruses' — burkov · 2026-10-03
- scaling01 Sparks Debate: Maybe We Should Hope Alignment Never Gets Solved — scaling01 · 2026-10-03