Hypothesis SubspacePaul Bricman12 Sep 2022 11:55 UTCA living collection of alignment proposals I’m exploring at Refine, a program hosted by Conjecture.Oversight Leagues: The Training Game as a FeaturePaul Bricman9 Sep 2022 10:08 UTC20 points6 comments10 min readLW linkIdeological Inference Engines: Making Deontology Differentiable*Paul Bricman12 Sep 2022 12:00 UTC6 points0 comments14 min readLW linkRepresentational Tethers: Tying AI Latents To Human OnesPaul Bricman16 Sep 2022 14:45 UTC30 points0 comments16 min readLW linkInterlude: But Who Optimizes The Optimizer?Paul Bricman23 Sep 2022 15:30 UTC15 points0 comments10 min readLW link(Structural) Stability of Coupled OptimizersPaul Bricman30 Sep 2022 11:28 UTC25 points0 comments10 min readLW linkCataloguing Priors in Theory and PracticePaul Bricman13 Oct 2022 12:36 UTC13 points8 comments7 min readLW link