Thread

Replies (8)

I've noticed that the people that marvel over AI's informative capacity have only asked AI questions that the human promptor does not know the answer to. 🤦🏻‍♂️ Other than one or two illustrations, I've only received bad data.
🛡️
Why would you use AI for a simple spell check :-) Translator yes is a good usecase , but translations never work because the language is not mere replacement of words .. even professional humans translators can't bring in the writer's vibe ..
🛡️
If you're writing a whole book and having AI check it for logical, semantic, or grammatical errors, you might as well include spelling checks too. But ChatGPT only catches about 40% of the errors and produces many false positives and negatives. Claude has actually very good translation capabilities, if you know how to prompt it.
🛡️
Absolutely not. I had 3 models write 1-2 chapters each, and it was all garbage. Logical contradictions, vague statements, etc. What you can use AI for, though, is to create a bullet-point outline of the plot after you've provided the general flow, and possibly check your story against it. That's called a "story bible" or something similar. And feeding the entire story each time isn't possible because their memory isn't sufficient. Eventually, it "forgets" the beginning of the story. And generally, after a while it just writes nonsense. But you can upload this 'story bible' each time you want sections checked.