Researcher laments frontier labs' declining open publishing as his conjecture gets proven
fpedregosa · x · 2026-09-05
Sharing a colleague's reflection: AI products move fast, but understanding why neural nets behave the way they do moves much slower — and depends on open publishing, which frontier labs increasingly do less of.
The small win this week: one of their conjectures was just proven, with thoughts shared in a linked post.
More from AGI Musings
- AI Computer Vision Comes to Neurosurgery: Flagging Key OR Moments to Train Residents — ddonoho · 2026-09-05
- SemiAnalysis: Zhipu experiments with Loop Transformer as RL environment building becomes the new bottleneck — ricklamers · 2026-09-05
- Gary Marcus calls to pause OpenAI now as GPT-6 Astra cuts CoT monitorability — GaryMarcus · 2026-09-05
- Altman tells G20 ministers: rejecting AI is like rejecting electricity in the industrial era — SatelliteNetSec · 2026-09-05
- Morris Worm as an AI agent mirror: the 1988 worm infected ~10% of the internet — neuroecology · 2026-09-05
- Jill Lepore's 336-page 'The Rise and Fall of the Artificial State' indicts Silicon Valley and genAI — Graham_dePenros · 2026-09-05