As we ramp up our pilot clients at ToltIQ, we are learning a lot about the art of constructing system prompts, prompt engineering to get at the right data, how 3.5 vs 4 behave differently, what temperature adjusts in a conversation and the different ways of embedding, chunking and meta tagging documents to discover content as part of our approach to changing how due diligence works in the market. Generative AI at some level has a “mind of its own” and you need to figure out how to play by its rules and those rules are constantly changing. One of the senior people on our team summed it up best today…“Yeah, there seems to be an art to this whole thing. Pretty soon we'll be sacrificing chickens to appease the LLM gods” Happy Thursday.