Every time I try to talk to ChatGPT4 about anything more complicated than doing madlibs with wikipedia and stack overflow entries, it goes off the rails immedia

Every time I try to talk to ChatGPT4 about anything more complicated than doing madlibs with wikipedia and stack overflow entries, it goes off the rails immediately. My sense is that LLMs are just a convenient interface to _certain kinds_ of knowledge that are well represented in the training data. I expect that generalizing to other domains for which training data is not available will prove very difficult. I must admit, the characterization of the responses of ChatGPT as "being like that of an expert" strikes me as utterly absurd. In essentially every field where I have expertise, talking to chatgpt4 is an exercise is confusion (and not on my part).