I’m working on a novel, and ChatGPT has been my companion / editor / fact checker the whole way. (Yes, of course I check AI’s facts.)
The quality of the interaction with ChatGPT is astonishing. It seems to understand the manuscript the way a human editor would. It gives me excellent feedback and suggestions on pace, character development, tone, the story arc …. You name it. If I was working with a human editor I couldn’t ask for much better.
I make it a point never to use its suggestions. The book is mine, not ChatGPTs. Still, the feedback is very valuable.
Don’t believe what the geeks say
Some people will tell you that all this apparent wizardry from the large-language models is simply a product of predicting the next most likely word. Don’t believe it. The level of analysis and, yes, understanding of the text goes way beyond that. There’s something very spooky going on in those energy-hogging circuits.
Does ChatGPT have a memory problem?
ChatGPT did a great job with my manuscript until I got to around 20,000 words. After that it started making silly mistakes, like giving me advice for a character I had already killed off, or giving suggestions for story arcs I had already precluded.
This was both a surprise and a disappointment because it was doing so well up until that point. I suspect there’s some memory limitation and it can’t handle text above a certain word count.
Tip: If you’re using AI in a professional application, check on that! Make sure the AI can handle the volume of text you’re asking it to process.
Grok to the rescue
A friend suggested that Grok was better than ChatGPT at this sort of work, and let me tell you — is it ever!
I was impressed by what ChatGPT could do, but I was astonished at the depth of analysis from Grok. And it didn’t have a problem with the longer text.
Is the feedback honest?
Well … either I’m a really good writer, or AI has a problem with flattery. While it will give me constructive feedback and criticism, it’s always couched within a lot of praise.
I actually am a decent writer, but it’s pretty clear to me that both the ChatGPT and Grok algorithms believe in buttering up the user. They may be following the logic of a former pastor who recommended that you sandwich criticism between two slices of praise. That may be more effective, and maybe that’s why AI does it, but it makes me wonder if you’re getting the straight scoop.
Still, the experience is amazing, and I’ve been very impressed with the feedback I’ve received from both systems.
Give it a try and let me know how it goes. And if you need help applying AI to your own business, give me a call.
