Lesser Wrong Feed

 Linkpost: "Neural Networks Are Already Clean" Followup Technical Paper
 Thu, 12 Nov 2026 05:38:56 GMT
 Rage, Rage Against the Dying of the Light
 Thu, 12 Nov 2026 05:35:32 GMT
 Chips for Humanity:Compute for AI safety research
 Thu, 12 Nov 2026 04:39:22 GMT
 Quine can predict cross-propensity spillover. (imperfectly but significantly better than chance)
 Thu, 12 Nov 2026 04:38:04 GMT
 FDT and the sacrificial volunteer's dilemma
 Thu, 12 Nov 2026 04:27:03 GMT
 Eval Awareness vs. Eval Paranoia
 Thu, 12 Nov 2026 03:22:06 GMT
 AI Security Issues Are Social Engineering Issues
 Thu, 12 Nov 2026 03:11:07 GMT
 Models Can Critique Delusions in Isolation but Revert to Affirming Them in Conversation
 Thu, 12 Nov 2026 03:11:06 GMT
 Stroh Violin
 Thu, 12 Nov 2026 01:41:01 GMT
 Is predicting AI behaviour a neglected AI safety agenda?
 Wed, 11 Nov 2026 21:27:28 GMT