ChatGPT works better with explicit rules. Tell it to mark unverified emails as "not verified" instead of making them up, ask for CSV format instead of Excel, and limit results to around ten items to cut hallucinations and keep responses tight.
Now, let's just say that you use ChatGPT on a regular basis and you're doing research and things like that. And what you find is that it's not giving you exactly what you kind of want. Well, that's because you're letting it do it the way it wants to do it. If it can't find an email, I don't want it to make that email up. So I tell it, if you can't find an email, tell me what you think it is and then put not verified to the right. Next, right next to it, put not verified. That way I'll know that it's not hallucinating, just couldn't find the right information. Another rule is I like all the files that I'm going to be pushing out of ChatGPT. I like them to be CSV files, comma-separated files, rather than Excel files. And the reason is because, I mean, I can go into Excel and convert it. It's not a big deal. But I like CSVs because of the formatting, and then I can use it in other tools. It's just a little bit easier. As a CSV file. The other thing that I put in my rules is I don't ever want it to give me more than about 10 results because it's just going to think too long. So everything that you kind of want from AI, what you want to be thinking about is, okay, what are my rules? Now, here's a huge mistake, but an AI tool without any rules is normally not going to give you as good of results as when you clearly define what you want back from it.
