wassname/awesome-ai-constitutions: AI constitutions and model specs, in date order, with quotes from studies on what generalises ·
AI constitutions and model specs: the written documents that say how a model should think and behave, and that are used to train it. The term comes from Bai et al. 2022: The idea is that human supervision will come entirely from a set of principles that should govern AI behavior, along with a small number of examples used for few-shot prompting. Together these principles form the constitution. Studies on what generalises (quotes): Oldest first. Unofficial reconstructions of unpublished documents are marked as such. * marks good starting points. Three Laws of Robotics From the short story "Runaround", quoting the fictional Handbook of Robotics, 56th Edition, 2058 A.D.: Sparrow rules 23 natural-language rules for a dialogue agent. Raters judged each rule separately, which trained rule-specific reward models. Rules are listed in the paper's appendix. Constitutional AI principles The paper that named the method: a model critiques and revises its own answers against 16 written principles, t