Mesa-Optimization

Applied to Alignment Problems All the Way Down by peterbarnett at 6d

Applied to Inner Alignment: Explain like I'm 12 Edition by elifland at 16d

Applied to Motivations, Natural Selection, and Curriculum Engineering by Oliver Sourbut at 1mo

Applied to Understanding Gradient Hacking by peterbarnett at 2mo

Applied to Some real examples of gradient hacking by Oliver Sourbut at 2mo

Applied to Feature Selection by Zack_M_Davis at 3mo

Applied to [Proposal] Method of locating useful subnets in large models by Quintin Pope at 3mo

Applied to Modeling Risks From Learned Optimization by riceissa at 3mo

Applied to Meta learning to gradient hack by Quintin Pope at 4mo

Applied to Towards Deconfusing Gradient Hacking by leogao at 4mo

Applied to Obstacles to gradient hacking by leogao at 5mo

Applied to Thoughts on gradient hacking by Ruby at 5mo

Applied to Approaches to gradient hacking by adamShimi at 5mo

Applied to Garrabrant and Shah on human modeling in AGI by Rob Bensinger at 6mo

Applied to Utility ≠ Reward by Steven Byrnes at 7mo

Applied to Thoughts on safety in predictive learning by Steven Byrnes at 7mo

Applied to Gradations of Inner Alignment Obstacles by Multicore at 9mo

Applied to Does SGD Produce Deceptive Alignment? by Mark Xu at 10mo

Applied to Formal Solution to the Inner Alignment Problem by Multicore at 1y