AI checking help

[Lifestyle66]

This is exactly the problem with AI as a beta-reader or doing a critique. No matter which text you ask the AI to process, it will always come up with show-dont-tell as advice and often in sections where showing just won't work. This means it just repeats the statistically most frequent writing advice. Alternatively, it might also recommend better transitions, even if it is obvious that the author wants an abrupt turn. So the AI will try to push you towards its own sloppy, overexplained style, with clichéed shows.

It's all in how you prompt it.

As I explained in my last post here, if the AI comes back with a point you disagree with due to the way you are evolving the story arc, you can tell it that is deliberate (as in "I want that misogynistic character for future use") and it then advises around that new known parameter.

The AI gives a dispassionate critique, and it doesn't argue. You need to teach it what YOU want to do with the story for it to give better responses.

And BTW, to head off the next objections, you are not training the AIs by feeding them your stories. The work involved in properly formatting the data upgrades, deconflicting the data parameters, testing the responses, and burning in the new info to the AI model takes extensive human reviews and testing. It took me HOURS and DAYS of writing, testing, and deconflicting the info on my immediate 12 family members to get correct responses from the local AI I'm building! Any particular random story is only useful to the humans who are doing that nug-work of formatting concise data for improving their AI models. And they can get your stories to read any time they want.

 

[Britva415]

just write the improvements yourself

This, all day long.

Now: I'm not recommending anyone ask AI for revision advice. What I'm saying is, if you do, then, take the advice and then go write the revisions yourself. Don't copy and paste from the output of a LLM and don't let a tool change the words in your draft document automatically.

Punch the keys for God's sake - type your own words yourself. That's still not a guarantee that you'll evade AI "detection," but it does eliminate one of the risk factors.

 

[Britva415]

You're conflating different goals in different stories aimed at different audiences to create a strawman argument.

It's not a strawman argument. It's exactly what you (said you) asked for: A critique from a feminist/lesbian point of view.

Too bad she's a human, eh?

 

[Lifestyle66]

It's not a strawman argument. It's exactly what you (said you) asked for: A critique from a feminist/lesbian point of view.

Too bad she's a human, eh?

She threw in the words "slut" and "misogynist" which were not part of my post, but part of that story!

Strawman!

"Let's fight about THAT, instead of the use of an AI tool in writing!"

I referred to that story to illustrate how an author can ask the AI for a specific perspective in it's review. You can feed it the same story and ask for a "general critique", "a lesbian POV", or "a misogynist POV", and you'll get different critical points in each to address if you want to make the story better accepted by THAT particular subset of readers.

You can't do that when asking a human beta-reader. They already have their own POV, take it or leave it.

EDIT: BTW, the AI did point out that using the word "slut" would be found offensive by a lesbian or feminist audience. And as I said, I instructed the AI as to WHY I was including it there for future use in shaping one of the FMCs.

 

[Britva415]

She threw in the words "slut" and "misogynist" which were not part of my post, but part of that story!

Strawman!

I guess I've lost the thread. I don't see how it's a strawman for something that's "not your story." You literally said you asked for a critique of the story. And that's what you got.

If anyone is strawmanning that for something else, it's...

 

[Lifestyle66]

I guess I've lost the thread. I don't see how it's a strawman for something that's "not your story." You literally said you asked for a critique of the story. And that's what you got.

If anyone is strawmanning that for something else, it's...

Correct, you lost the point of the thread.

This thread was about using an AI as a writing tool, not about any particular offensive words in a story. So,, to conflate the content of one story with the use of an AI as a tool is deflection and a strawman argument.

"Let's get the bull charging at the strawman, so it will forget about the AI as a tool."

 

[nice90sguy]

Here's another example of useful AI proof-reading:

I asked an AI to look for Britishisms in both dialog an narrative in one of my stories, and it found a ton of them (well, dozens), most of which were new to me. And it provided a shortlist of the most imptrant ones too, for each chapter.

1. pinking → pinging / knocking / ticking as it cooled
2. basement car park / underground car park → parking garage / underground garage
3. apartment block → apartment building
4. decking → deck
5. trained as a lawyer → went to law school / I’m a lawyer
6. It takes all sorts → Takes all kinds
7. fetch → get / go get, except where deliberately quaint
8. Ford Mercury → Mercury
9. Dime Box, Tx → Dime Box, Texas / Dime Box, TX
10. Reduce Anna’s phonetic spellings: di’nt, complainin’, li’l, etc.

Bottom line: the Jimmy dialogue itself is solid. The American verisimilitude problem is not his voice; it is the surrounding automotive and environmental vocabulary. The bigger dialogue fixes are Gina’s British legal idioms and Anna’s slightly over-visible dialect spelling.

That's what it found in the last 3K words of the story. The first two chapters had even more issues.

Before publshing the story here back in 2021, I asked the late great TxRad to beta read that last chapter for me. He gave me a few tips, but obviously he was nowhere near as patiently methodical as ChatGPT, and found only a couple of issues.

I hope my posts here about how to get the best out of AI and avoid its pitfalls might convince some ppl that it's worth the effort to add it to your creative writing arsenal. But I repeat my warning. Treat it with sceptisim, and NEVER ask it to be creative, only to be analytical and critical, whch it excels at.

I think that the better the writer you are, the more you'll benefit from using AI in this way.