Rob LeClerc argues training-time models need no perfect guardrails, citing 15-min monitoring catch
robleclerc · x · 2026-10-05
A debate on X over AI safety during model training. Rob LeClerc argues production-grade protective layers aren't needed for models still in training: based on years evolving neural networks, he says perfect one-shot conformance is unrealistic since neural nets are inherently slippery to control, and mistakes are how we learn. If desired, a dozen specialized models could be trained to shut down out-of-policy actions, but over-monitoring isn't cost-effective while risks remain benign. He notes a general monitoring AI caught the issue 15 minutes later. Neel Nanda counters that failing to prevent harm in training is still a disaster.
More from Safety
- Carnegie Analysis: China Meets AI Slowdown Calls With Skepticism, Not Rejection — mattsheehan88 · 2026-11-20
- New House AI Safety Caucus one-pager was apparently designed by Claude — NinaDSchick · 2026-11-20
- Tech outages giving Y2K vibes, AI-agent style: experts see something deeper — TechCrunch AI · 2026-11-20
- Senator Van Hollen calls for mandatory third-party AI evals and a ban on recursive self-improvement — peterwildeford · 2026-11-20
- Model Fight Card 16: Tenet models dominate jailbreak arena with 21/24 escapes — Dave_Maynor · 2026-11-20
- Prime Intellect Endpoint Served Refusal-Stripped DeepSeek Copy: Refusals Fell 98%→1% — matthew_d_green · 2026-11-20