\n<\/p>\n
Ever wonder why conversational AI like ChatGPT says \u201cSorry, I can\u2019t do that\u201d or some other polite refusal? OpenAI is offering a limited look at the reasoning behind its own models\u2019 rules of engagement, whether it\u2019s sticking to brand guidelines or declining to make NSFW content.<\/p>\n
Large language models (LLMs) don\u2019t have any naturally occurring limits on what they can or will say. That\u2019s part of why they\u2019re so versatile, but also why they hallucinate and are easily duped.<\/p>\n
It\u2019s necessary for any AI model that interacts with the general public to have a few guardrails<\/a> on what it should and shouldn\u2019t do, but defining these \u2014 let alone enforcing them \u2014 is a surprisingly difficult task.<\/p>\n If someone asks an AI to generate a bunch of false claims about a public figure, it should refuse, right? But what if they\u2019re an AI developer themselves, creating a database of synthetic disinformation for a detector model?<\/p>\n What if someone asks for laptop recommendations; it should be objective, right? But what if the model is being deployed by a laptop maker who wants it to only respond with their own devices?<\/p>\n AI makers are all navigating conundrums like these and looking for efficient methods to rein in their models without causing them to refuse perfectly normal requests. But they seldom share exactly how they do it.<\/p>\n OpenAI is bucking the trend a bit by publishing what it calls its \u201cmodel spec,\u201d a collection of high-level rules that indirectly govern ChatGPT and other models.<\/p>\n There are meta-level objectives, some hard rules and some general behavior guidelines, though to be clear these are not strictly speaking what the model is primed with; OpenAI will have developed specific instructions that accomplish what these rules describe in natural language.<\/p>\n