Rob Leclerc: model unmonitorability is a lab choice, not an inevitability
robleclerc · x · 2026-09-18
Responding to complaints that models are becoming unmonitorable, Rob Leclerc argues that training for form over function effectively relaxes constraints on legibility — so the trend should be expected. But he pushes back on framing it as inevitable: adding legibility constraints would increase legibility at some cost to model performance. He calls it dangerous to imply that as models get more powerful, opacity is simply the way things will be. It's a design choice wholly under the control of the labs, not a law of nature.
More from AGI Musings
- Researchers argue p(doom) launders evidence-free fears through pseudo-quantification — RexDouglass · 2026-09-18
- Researcher chose genetics over AI in 2020, citing an unimaginable superintelligence future — erikphoel · 2026-09-18
- How Yudkowsky and Bostrom shaped today's AI safety discourse 12-16 years ago — Justin_Halford_ · 2026-09-18
- The intelligence loan: frontier labs subsidize your access and repay themselves by distilling your edge — cjmaddison · 2026-09-18
- The knowledge economy is over: Dan Shipper on the 'Allocation Economy' where every maker becomes a manager — danshipper · 2026-09-18
- The 'Claude Coworker' does all his work unchecked—and the AI now outperforms him — omooretweets · 2026-09-18