FYI - I’m thinking on how to sketch out ideas in a “mind map” (Kinopio) and have GPT ghost-write a fleshed out post from that (this would also be good for README.md files in dev repos).
I had some success with jamming Kinopio-generated JSON into KAGI Summarizer, but hit a road-block when I tried to pour the same JSON into ChatGPT-3 (prompt too long, and, won’t accept URLs).
I’m thinking of trying again with Llama-2, but, thus far haven’t gone down that learning curve...
I, also, have had success with writing a terse paragraph, then having ChatGPT turn it into a chapter for a “book”. Prompt: ‘edit this chapter and flesh it out ’ <followed by the terse paragraph text>), result: successfully generated chapter(s), but, lots of reading for me. Approving, editing, etc. - stuff I don’t really want to be bothered with. I will gladly share, but, it’s too long to post here.
It looks like feeding short articles / chapters to ChatGPT will work, but I wish for something even simpler. I guess I could dump my thoughts into Kinopio, then dictate chapters into text files while prompting myself by looking at the Kinopio thoughts and recording my ramblings (say, using speech-to-text). If my other experiments fail, I will fall back to this strategy.
Along the way, someone mentioned ‘
type.ai’ to me. It looks interesting but I haven’t had time to check it out.
An experiment with
type.ai managed to produce a reasonable summary of the same Kinopio thought-map as above, but, the experimenter typed the text for each thought-bubble in manually.
I would enjoy hearing/reading about any experiments you try. In fact, if you are successful, you can have GPT generate a blog post about how you did it :-).