4 Comments
User's avatar
Les Barclays's avatar

Great post on *why* AI writing is subpar!

I’ve realised that when I read AI generated content out loud and then read human generated content, there’s a difference in cadence/flow and doesn’t seem to flow as well as it would if it was human written (assuming said person can write well).

Kent Myers's avatar

I am using Gemini in novel writing. I give it a short draft with all the ideas, and it spits it back with some more staging, repetition of motifs, and dialogue that sounds like my characters. I consider the two, together, as a first draft. When I rewrite, I think there's never a Gemini sentence that remains untouched. I have never been satisfied with a first draft of anything, so I have no expectations for this first draft. As an old professor said, "The best writing is re-writing." I asked Gemini to compare itself to Claude, and it says that Claude writes with more power and concision, but it says it gives me more to work with as a re-writer, and I agree. As it offered in conclusion, after pointing out that I came up with all the good ideas and all it did was string it out "We are a great team!" I don't expect it to give me finished work, but it does prompt me to add a lot of points that I want to add, in my own way. I don't mind people criticizing me for using the boost, but I do mine people who say I'm not a legitimate writer, and that the product is not mine.

Jerry M. Withers's avatar

I also agree that the best writing is rewriting, in fact, typically over several if not many rounds thereof.

Dorian's avatar
5dEdited

Very true. AI models fail at:

1. Maintaining a consistent, well-structured idea or line of thought in long-form writing.

2. Randomly creating repetitive short sentences, like a machine glitching.

3. Maintaining a consistent writing style, and in longer pieces, the logic starts to fall apart.

4. Having precise control over word choice, meaning, and the thesaurus.