Im testing ways to explain the overhype issue with GenAI. The current iteration I have is harkening to Coke vs Pepsi. LLMs, like Pepsi test incredibly well wit

Im testing ways to explain the overhype issue with GenAI. The current iteration I have is harkening to Coke vs Pepsi. LLMs, like Pepsi test incredibly well with the first use. It’s a different story once you start dealing with it long term, hallucinations being a core issue ( feature, not a bug - so its not going anywhere) With that said, there is 1 place that it will work well: Code. Since words are law and product, this is a straight productivity boost. The other place it should have an impact is law, however nothing akin to a law copilot is making the rounds. I’m curious why. Is there simply more code available to train on that cases? Does the word length of legal documents break context windows?