Episode 4 - Breakthrough

My biggest problem with content generated by ChatGPT is that there is no way I can make ChatGPT remember my edits. If that were possible, then the output would automatically improve with each piece of content it creates for me, or in ChatGPT-speak, with each conversation. But instead, ChatGPT only “remembers” my edits indirectly, lumping them in with millions (if not billions) of edits from everybody who engages with the platform.

?

As it turns out, this may not be a problem after all. Here’s the newest trick I learned: If you have an account, the opening screen presents you with a small button in the lower left corner that

has your initials on it. When you click the button, ChatGPT offers several options, including “How would you like ChatGPT to respond?” For my most recent experiment, I pasted in several hundred words of a previous article I’d written and told ChatGPT to use that text as a style model.

?

The result: a perfectly acceptable draft that I could show to a client with only a few minor edits. I haven’t gone further in this direction yet, but I suspect that I could get similar results pasting in copy from a target publication or site. For example, I think it would be possible to paste in a portion of an article from, say, Fortune or CIO and get a result that was stylistically appropriate. I’ll let you know.