Using AI as a reviewer to improve my writing

[Bramblethorn]

You were made by just two strands of nucleotides composed of nitrogen-containing bases: adenine, guanine, thymine, and cytosine.

This isn't even accurate as a statement about genetics, let alone in the broader sense that you're trying to apply it here.

 

[Cacatua_Galerita]

...the rest are quite predictable. For instance, I knew exactly how and who would react.

Possibly, but I doubt you knew what words they would use. (But feel free to preemptively post my next reply, if you can predict it.)

I think a lot of the free-will/predictable-behaviour debate is a matter of perspective. We have a self described ‘mystic’ conversing with a bunch of ‘IT geeks’. The concept of ‘predictable’ is different between these groups.

A psychologist once explained to me that they screen IT people differently. If asked ‘have you ever thought about suicide?’ the ‘normal’ person would answer no, because they’ve never seriously considered it. The IT geek would answer yes, because they had to spell the word in a spelling bee once, therefore they had thought about it. IT folk are very boolean-logical, black and white (true/false) thinkers. They will choose pedantic correctness over colloquial conformance because programming computers has taught them to think that way.

I was agreeing with both sides in an earlier post because I think people are predictable (like the climate) but individuals aren’t entirely predictable (like the weather). The brain is a chaotic system, dependent on the quantum level interactions of sub-atomic particles. Hence the ‘chaos’.

There seem to be a bunch of paywalled papers out there about this:

Chaos and free will.

The Constraint of Chaos: Rethinking 'Freewill vs. Determinism'

What is chaos and how is it relevant for philosophy of mind?

And I’ll add this one for the ‘username checks out’ prize because I love the idea of a guy named “Neuringer” doing brain research:

Approximating Chaotic Behavior

(Could someone with an account ask ChatGPT to summarise these papers? )

 

[SmokingFap]

I’m not really all that interested in a computer program’s opinion of the emotional depth of my writing.

Thank you!

 

[XerXesXu]

(Could someone with an account ask ChatGPT to summarise these papers? )

I have the impression that a lot of researchers ask ChatGPT to write the abstract.

 

[Cacatua_Galerita]

A fleeting electrical glitch might activate a memory cell, and suddenly you’re craving a donut

Or in other words, a butterfly flapping its wings in Brazil can cause a tornado in Texas. I.e. A dynamic system that is highly sensitive to initial conditions.

Is the fleeting electrical glitch 'predictable'?

 

[Plathfan]

Or in other words, a butterfly flapping its wings in Brazil can cause a tornado in Texas. I.e. A dynamic system that is highly sensitive to initial conditions.

Is the fleeting electrical glitch 'predictable'?

If it follows a recurring pattern of occurrences, then yes.

 

[secretLace]

In the end it is a tool that some see use in. I use it to brainstorm and then evaluate its output. If you review it criticaly you can learn but a lot of it is generic and flawed in execution. Like some already pointed out, if you have a clear understanding of your capabilities then you might not need it but if you struggle then grammar och structure pointers could be useful. As useful as reading a book on grammar and structure and then apply it to your work.

Just realise that its not always right.

 

[Voboy]

Amid all the concerns about AI companies mining everyone's creations to fuel their programs, you want to voluntarily feed it yours?

This.

Fuck AI. I want nothing to do with it. My stories stand or fall based on my thoughts, my ideas, my self-critique. Otherwise, they are not wholly mine.

The handful of times I've ever taken any ideas or inspiration from anyone else, however niggling, I've made sure to acknowledge their help in a note at the top of my story. I think that's the only ethical way to take help in a published work: to tell the world about that help.

Do you publicly acknowledge the help this AI gives you, OP? Do the readers know they're reading something that's been approved by a computer?

 

[Voboy]

But that is because you are an exceptional writer. You don't need this. The way you create characters with depth is amazing. I fed some of your text into the AI to understand how you do it.

I've not read the whole thread, OP, but in case nobody else has pushed back on this?

Don't be feeding our work to your AI friend. Ever. It's not your work to experiment with in this way. You have a mind that is FAR superior to any AI; if you let yourself learn from other people's prose, you'll do much better using your own brain. And, on top of that, you won't be feeding other writers' work to a technology many of us find highly repugnant.

Do as you please with your own work; don't be messing around with anyone else's.

 

[Lifestyle66]

This.

Fuck AI. I want nothing to do with it. My stories stand or fall based on my thoughts, my ideas, my self-critique. Otherwise, they are not wholly mine.

The handful of times I've ever taken any ideas or inspiration from anyone else, however niggling, I've made sure to acknowledge their help in a note at the top of my story. I think that's the only ethical way to take help in a published work: to tell the world about that help.

Do you publicly acknowledge the help this AI gives you, OP? Do the readers know they're reading something that's been approved by a computer?

Have you tried it?


I just tried feeding ChatGPT with my latest work-in-progress, asking for a "review of a short story" (I'm a geek, and want to try staying abreast of the technology.) I was interested in seeing what type of feedback I would get.

I must admit, that although I don't want computer AIs doing everything, it did provide some VERY constructive feedback!

My wife listened to my story and merely said it's one of my better works, but provided no input as to content shortfalls or ways to improve it.. I have sent a copy to another author here as a beta-reader, and awaiting feedback (we're all busy, and it takes time.)

But in literally one second, ChatGPT sent back a one-page review, both praising the content as written AND pointing out several areas for improvement. One of its points in particular was a GLARING omission on my part, in that one of the key characters (the cheating husband) is "one dimensional" (that's Chat's exact critique) and needs further character building to explain his behavior.

Now, I won't ask the AI to do any editing or line-by-line critiques with recommendations, because that's going too far. But as a beta-reader, ... I have to admit that in this one test case, it both saves me time and provided constructive feedback RAPIDLY. So, I might look at it as a tool (when used sparingly) similar to "spell check", a tool to help me improve MY writing skills.

 

[Voboy]

Have you tried it?

No.

I have faith in my brain. My sentences say what I want them to say. When they don't? I change them until they do.

I have no idea why I would ever benefit from AI taking on that role. On the contrary: it would reduce the authenticity of my stories (something I care about quite a bit), it would make me a less attentive and more complacent writer, and it would probably reduce my critical thinking abilities.

I see no upside and plenty of downside.

My writing skills are MINE, not a computer's. I take pride in that. It's what makes me a writer: I do the writing. My stories come from nobody but me. I don't see why I'd want it any other way. When I can no longer do that? Then I'll stop writing.

I think the key to any creative endeavor is that the creator does the creating. Nobody (and nothing) else. I accept that others will disagree, and I guess that's fine. But if they can't express that disagreement (nor anything else) without a computer's help, then I don't value their opinion one bit. Because it's not theirs.

 

[Lifestyle66]

No.

I have faith in my brain. My sentences say what I want them to say. When they don't? I change them until they do.

I have no idea why I would ever benefit from AI taking on that role. On the contrary: it would reduce the authenticity of my stories (something I care about quite a bit), it would make me a less attentive and more complacent writer, and it would probably reduce my critical thinking abilities.

I see no upside and plenty of downside.

My writing skills are MINE, not a computer's. I take pride in that. It's what makes me a writer: I do the writing. My stories come from nobody but me. I don't see why I'd want it any other way. When I can no longer do that? Then I'll stop writing.

I think the key to any creative endeavor is that the creator does the creating. Nobody (and nothing) else. I accept that others will disagree, and I guess that's fine. But if they can't express that disagreement (nor anything else) without a computer's help, then I don't value their opinion one bit. Because it's not theirs.

All good points.

But how many of you who agree TAUGHT yourself how to write, with absolutely NO FEEDBACK from teachers, professors, editors, or beta-readers?

As I tried to point out, the AI (as I used it) gave me such constructive critique as I might get from a college professor. It effectively "graded" my submission as a "B" with written notes in red saying "a secondary character is one-dimensional and needed more explanation of his motives for agreeing to the wife's RAAC."

 

[Britva415]

just because it says it can't access the site doesn't mean it isn't accessing the site. This is likely one of the ring fences put up by the developers to try to prevent model poisoning. There will be massive ETL processes running in the background to feed the basilisk.

It can’t access the site. Whatever material ChatGPT was trained on, it happened in the past. The developers feed it new training material for each new version of ChatGPT, and newer Lit stories probably make up some of that material, but Chat GPT can’t just go browse the web.

 

[alohadave]

It can’t access the site. Whatever material ChatGPT was trained on, it happened in the past. The developers feed it new training material for each new version of ChatGPT, and newer Lit stories probably make up some of that material, but Chat GPT can’t just go browse the web.

Something that I think people don't understand is that training a model is very computationally intensive and is not done on the fly. The data used for a particular model is fixed in time.

They may capture data entered for use in future updates.

 

[XerXesXu]

It can’t access the site. Whatever material ChatGPT was trained on, it happened in the past. The developers feed it new training material for each new version of ChatGPT, and newer Lit stories probably make up some of that material, but Chat GPT can’t just go browse the web.

The LLM base model bots can trawl the site, with or without permission, but the user front end, ChatGpt etc, cannot.

 

[Britva415]

The LLM base model bots can trawl the site, with or without permission, but the user front end, ChatGpt etc, cannot.

The LLM base model behind ChatGPT isn’t the same one which is being updated by bots. Developers train the development version (copy, instance) of the LLM after it ingests new material, and release the trained version as an update to the live instance of ChatGPT, with a new version number.

It’s not a real-time process, it’s not even an “automated but delayed” process.

 

[XerXesXu]

The LLM base model behind ChatGPT isn’t the same one which is being updated by bots. Developers train the development version (copy, instance) of the LLM after it ingests new material, and release the trained version as an update to the live instance of ChatGPT, with a new version number.

It’s not a real-time process, it’s not even an “automated but delayed” process.

The base model is like a trained memory. It's frozen in time for the purpose of producing a useable front-end.

There's always a base model in training.

 

[Voboy]

All good points.

But how many of you who agree TAUGHT yourself how to write,

None, but that goes for every other creative endeavor too. Most of my own learning about how to use the English language came not from teachers, but from my own reading.

with absolutely NO FEEDBACK from teachers,

In all honesty, my ELA teachers had very little to teach me about creative writing. They all said as much, from an early age. I took what I could use and ignored the rest.

professors,

None. My university did not really believe in helping students learn to write; their position was that you ought to know it before you arrived.

editors,

As I've pointed out here before, I did work with an editor back when I was doing commercial erotica. She doubled as my publisher, so I took her suggestions because she was putting up money and, thus, taking on risk. The byline was mine, but the product was hers. We worked well together, and she liked me because I needed so little editing.

or beta-readers?

I don't use them. I did once, accidentally, and it was a strange experience for me. I did consider the suggestions, but only because the accidental beta-reader was a fellow writer who I respect. I didn't change anything, though.

As I tried to point out, the AI (as I used it) gave me such constructive critique as I might get from a college professor. It effectively "graded" my submission as a "B" with written notes in red saying "a secondary character is one-dimensional and needed more explanation of his motives for agreeing to the wife's RAAC."

I mean, that's cool and all. But could you not have summoned enough detachment to step back and do that process yourself? If not, that's intriguing to me; I can't really relate to that.

 

[ElectricBlue]

All good points.

But how many of you who agree TAUGHT yourself how to write, with absolutely NO FEEDBACK from teachers, professors, editors, or beta-readers?

I'm like @Voboy. Zero feedback from any teachers in terms of creative writing, except maybe in primary school. That hardly counts for erotica.

University taught (Arts Degree), but that was English Lit critique and history, not creative fiction. A year or two with two beta readers here on Literotica, but their input was more guidance, and just as often ignored. Decades as a business writer, but again, that's not creative, certainly not fiction (but does involve making a lot of shit up).

Did I ever set out to "teach myself to write?" No, I just started writing and learned as I went along.

I occasionally wonder what AI content might look like if I trained it only my own content, but I've not gone looking for a tool that does that. That would be interesting, I reckon, given how many themes and tropes are repeated.

 

[Lifestyle66]

Did I ever set out to "teach myself to write?" No, I just started writing and learned as I went along.

That's what I have done as well. I never took any creative writing courses and actually found high school writing in English to be tedious. But I started just over the last four plus years writing here and taking the feedback to realize what would have made a better story.

I don't consider it "teaching" myself. But I haven't just thrown shit against the wall to see what sticks. I've considered the comments and even the rating fluctuations, as well as the verbal comments from my wife (when she reads them) to adjust my writing the next story.

I don't claim to have some God-like prescience to know when I started what made a good story or believed I've covered all bases. I did learn that it was the subtle inputs from others which guided me to write somewhat better stories. So, I see the AI as yet another potential tool, like a faster/instantaneous beta-reader.

I might use it as a "wheel" to move me along to write a better story, while no allowing it to drive me. I think it's like recognizing that the wheel was not a bad invention. It just doesn't replace the lever for heavy lifting.

 

[Andreas_Kreuz]

Do you publicly acknowledge the help this AI gives you, OP? Do the readers know they're reading something that's been approved by a computer?

Since I have not yet published a story with this kind of AI help, it has not yet been an issue.
But reading your post, I realized I should put it into a note about the story.
I will be curious if Laurel considers that as a reason to reject the story due to AI use.

But my intention was not to use AI to write a single better story, but to practice writing in general.
So the outcry about the lack of emotional depth has helped me starting to think about emotional depth. What is it, how do I convey it. Do I actually want it? So my test story was called "A whore for one evening". In that setting I did not want the protagonist to ponder her emotions all the time, but rather take her through the motions emotionless.

With AI I can streamline my writing process. E.g. with such a prompt:
"Read this erotic adult story. Focus on the character of X. Collect all statements about her character and summarize it. Is she portrayed consistently? Do you notice inner conflicts? How are they handled.?

And I must say AI gives me answers. They are like the general writing guidelines posted here and on other sites, but directly applied to my text.

 

[StillStunned]

But my intention was not to use AI to write a single better story, but to practice writing in general.
So the outcry about the lack of emotional depth has helped me starting to think about emotional depth. What is it, how do I convey it. Do I actually want it? So my test story was called "A whore for one evening". In that setting I did not want the protagonist to ponder her emotions all the time, but rather take her through the motions emotionless.

And this is one of the reasons why an AI review is pretty useless. Depth of character is really only necessary if that's what the story is about. Plenty of short stories work perfectly well without it. They might focus on a single moment's experience, or an event, or a location, and the character's reaction to it. And that's even without considering that erotica generally focuses on a sexual experience, not an emotional one.

Like someone mentioned upthread, ChatGPT is aping what it's read in other reviews, without considering whether any of the factors are relevant.

For instance, I wrote a story called Fairytale of New York. Only 1.8k words, with minimal dialogue, and that very sparse. And yet according to the comments it's by far the most profound story I've ever written. Would ChatGPT get that?

 

[Lifestyle66]

For instance, I wrote a story called Fairytale of New York. Only 1.8k words, with minimal dialogue, and that very sparse. And yet according to the comments it's by far the most profound story I've ever written. Would ChatGPT get that?

Try feeding your story to ChatGPT to find out what it says.

I receive comments from readers, critiques from beta-readers, and had editors provide guidance in the past. And I SOMETIMES even take their recommendations to improve a story. But I've received some which I found went far astray from what I intended.

As I said above, when used sparingly, the AI review appears to be a very fast TOOL to use among others.

 

[StillStunned]

Try feeding your story to ChatGPT to find out what it says.

No thanks. It's my story, and I'm not giving any tech companies permission to use it.

 

[Voboy]

Try feeding your story to ChatGPT to find out what it says.

Why would he?

He wrote the story. He knows it.

This is primarily what bugs me about you pro-AI people: you're selling yourself short. Everything AI can do, in terms of probing your story, evaluating your characters, or critiquing your style, is stuff you can already do yourself. Your brain is fully capable of all this; indeed, it evolved specifically TO do all these kinds of critical thinking. And you're not letting it.

You keep your brain strong and vital by using it. My overriding concern about tech in general, and certainly about AI for creative writing, is that it makes us underreliant on that marvelous thing we all have between our ears.