AI for retrospective story/author analysis

[SomaSlave]

Since I've been using Claude, I've been impressed with its analytical skills, so I set it to work analyzing my evolving writing style and my current large story arc. For the record, I never use AI to create a story, only to check it for continuity and grammar/spelling issues. Any other analysis takes place after it's been published.

It did a decent job identifying a story arc and character development for the 15 installments I've completed so far. It did a decent job with that and created a story bible for me that will be helpful in me catching any continuity issues in the next few installments.

After that, I gave another task. I have a series of stories here that recount my own explorations in only slightly embellished form. These were written many years ago, when I was looking for my voice. Since my current arc is based is essentially a fictional "what-if" extension of those explorations, I thought it would be interesting to see if Claude could identify any influences of those earlier tales in my current set. I gave it the background and told it to treat them as fiction in its analysis.

It was brutally honest about my flaws as a writer back then (yes, it was possible for me to be a worse writer than I am now), but also traced my improvement in writing as the stories progressed. Then came the stunner: while reviewing the eighth story, it said, "This character is Jim," referring to one of the leading characters in the current series. I had no idea: as far as I was concerned, Jim was an original creation for the new stories. Claude noted the description of the man in the much older story was almost identical to Jim. Not only that, his character and actions mirrored Jim's to a disturbing degree.

Somehow, I had been subconsciously that man and that encounter when I was working on The Shopping Spree. That epiphany has forced me to reexamine the series in that light. The literary analysis was good. The connection it made was profound. I'm passing on as an interesting use of AI, particularly Claude, for after-the-fact work.

 

[AG31]

Do you submit the whole text in your prompts? I've had conflicting responses from ChatGPT about whether it can actually read stuff on Literotica.

 

[TheWritingGroup]

Since I've been using Claude, I've been impressed with its analytical skills, so I set it to work analyzing my evolving writing style and my current large story arc.

Be aware: generally, that means "Claude is sycophantic and is agreeing with me."

 

[SomaSlave]

Do you submit the whole text in your prompts? I've had conflicting responses from ChatGPT about whether it can actually read stuff on Literotica.

I gave it the entire story from my own repository. I was pleased that it stayed away from the guardrails.

 

[SomaSlave]

Be aware: generally, that means "Claude is sycophantic and is agreeing with me."

I've added instructions in the default settings to specify I'm looking for pushback, not affirmation. It should know after many conversations that I do not like sycophancy...and it was quite willing to be quite critical of my early writing.

 

[TheWritingGroup]

I've added instructions in the default settings to specify I'm looking for pushback, not affirmation. It should know after many conversations that I do not like sycophancy...and it was quite willing to be quite critical of my early writing.

I don't know about Claude specifically, but the LLMs I've read about forget everything the instant you end a conversation. You have to repeat every single priming thing every single interaction.

 

[SomaSlave]

I don't know about Claude specifically, but the LLMs I've read about forget everything the instant you end a conversation. You have to repeat every single priming thing every single interaction.

In this case, I've set up project and uploaded each story. That keeps them in memory and I can return to that same conversation without missing a beat. Sonnet is a very good LLM and well worth the Pro subscription...and it nudges me when it's time to stop typing and go to bed.

 

[HeyAll]

I use Grok

It's a fantastic assistant for sure.

The spell check needs improvement though.

Grok was able to catch several typos, but Grammarly could catch more.

 

[Five_Inch_Heels]

NO fake nonsense. Not Ever. Not at all.

I almost wish they'd start banning people from here for even mentioning any version of it.

 

[PennyThompson]

I've added instructions in the default settings to specify I'm looking for pushback, not affirmation.

Okay, but that's still sycophancy... It's just doing its best to generate output that statistically matches the given input.

You could ask it to critique Bukowski every time he's florid and overly verbose, and it will make up examples to support the prompt.

 

[EmilyMiller]

Okay, but that's still sycophancy... It's just doing its best to generate output that statistically matches the given input.

You could ask it to critique Bukowski every time he's florid and overly verbose, and it will make up examples to support the prompt.

People do not understand the principles behind LLMs…

 

[anthrodisiac]

Asking it to critique your work results in it giving you a critique because you ask for it, not because the work is actually worthy of critique. And the critique is usually nonsensical and beyond wrong. Or mildly correct, but the prescribed solution is idiotic. Shockingly, statistical prediction machine isn't great at actual thoughtful analysis of work where it doesn't already have a ton of examples of critiques it can offer. Who knew?! (Answer: A lot of us.)

The only good use of AI is for me to show it something I wrote that I'm struggling with, ask it what it thinks, then proceed to tell it why it's wrong. It's not unlike asking someone who's read four books in their lifetime, three of them children's picture books, what works about the handful of paragraphs, and then you have to explain to them why nothing they said makes any sense. It's actually been kinda useful at times when I'm not quite sure why something isn't working, I can't figure it out at all, and having to explain it to an idiot at least puts me in teacher mode in a way I wish I could just do with myself. I don't really have anyone to bounce ideas off of, so this is the very mediocre solution I've come up with a couple times when I'm at my wit's end.

Not that explaining why it's wrong is useful in any way, but I get to organize my thoughts and response, then roll my eyes when it goes, "Yeah, that makes perfect sense! Gosh, you're so smart, and such a pretty archaeopteryx!"

Bitch, I know

 

[Lifestyle66]

I don't know about Claude specifically, but the LLMs I've read about forget everything the instant you end a conversation. You have to repeat every single priming thing every single interaction.

Not if you keep the browser window open.

You can go back to the same conversation with the AI and it will retain some of the earlier info. Also, creating an account and logging back in allows you to continue the same conversation.

But AI's have their limits with the amount of text you feed them to review. Anything more than about 5K words can have some of the details "forgotten" or ignored in its final review of your story.

EDIT: I'm building my own home AI on a new graphics workstation (i9 processor with Nvidia Blackwell Pro 2000 GPU, along with an older VR computer for running Home Assistant). And in asking Gemini for what to do about software errors integrating the two systems, it gives me things to try. But I can go back days later, and it remembers the goals of my project and the hardware I'd previously described.

 

[EmilyMiller]

Asking it to critique your work results in it giving you a critique because you ask for it, not because the work is actually worthy of critique. And the critique is usually nonsensical and beyond wrong. Or mildly correct, but the prescribed solution is idiotic. Shockingly, statistical prediction machine isn't great at actual thoughtful analysis of work where it doesn't already have a ton of examples of critiques it can offer. Who knew?! (Answer: A lot of us.)

The only good use of AI is for me to show it something I wrote that I'm struggling with, ask it what it thinks, then proceed to tell it why it's wrong. It's not unlike asking someone who's read four books in their lifetime, three of them children's picture books, what works about the handful of paragraphs, and then you have to explain to them why nothing they said makes any sense. It's actually been kinda useful at times when I'm not quite sure why something isn't working, I can't figure it out at all, and having to explain it to an idiot at least puts me in teacher mode in a way I wish I could just do with myself. I don't really have anyone to bounce ideas off of, so this is the very mediocre solution I've come up with a couple times when I'm at my wit's end.

Not that explaining why it's wrong is useful in any way, but I get to organize my thoughts and response, then roll my eyes when it goes, "Yeah, that makes perfect sense! Gosh, you're so smart, and such a pretty archaeopteryx!"

Bitch, I know

As you and I know, LLMs produce textual answers based on what it has been trained on and the probabilities that certain outputs meet certain inputs.

This is great if you say: ChatNazi, does the Earth go around the Sun or the Sun go around the Earth?

The LLM has read many many texts answering that question. Middle school essays saved on Google Classroom, text books, pop-science books, articles in science magazines, blog posts, half of Quora… At least most of the time, it will regurgitate something semi-sensible, as we are talking about settled fact.

If you say: ChatNazi what do you think of my latest erotic story?

Then the LLM has nothing to go on. The New York Times has yet to review your work. What it does have is generic shit that appears in story reviews - regardless of the actual story. And, as its programmed to fake it when it doesn't have a good match, it just makes stuff up, interspersing a few details from your text with a review of Harry Potter.

And people fall for it.

 

[Voboy]

In this case, I've set up project and uploaded each story. That keeps them in memory and I can return to that same conversation without missing a beat. Sonnet is a very good LLM and well worth the Pro subscription...and it nudges me when it's time to stop typing and go to bed.

So... you need a computer to tell you what you've asked it to tell you critique your own story, and you also need it to remind you you need to go to bed.

Are you even a functional human being at that point? Does the LLM wipe your ass for you as well?

The degree to which so many of my fellow humans are gleefully throwing their brains down the toilet so that they can cede control to a machine is never going to stop bothering me.

 

[EmilyMiller]

So... you need a computer to tell you what you've asked it to tell you critique your own story, and you also need it to remind you you need to go to bed.

Are you even a functional human being at that point? Does the LLM wipe your ass for you as well?

The degree to which so many of my fellow humans are gleefully throwing their brains down the toilet so that they can cede control to a machine is never going to stop bothering me.

Part of the fun and challenge of writing is all the things that LLMs claim to do and are awful at. And any human will be a vastly superior foil for your ideas and conundrums.

I don’t know if it’s because some of us are ashamed of writing erotica that we fail to reach out to other humans about it.

 

[anthrodisiac]

As you and I know, LLMs produce textual answers based on what it has been trained on and the probabilities that certain outputs meet certain inputs.

This is great if you say: ChatNazi, does the Earth go around the Sun or the Sun go around the Earth?

The LLM has read many many texts answering that question. Middle school essays saved on Google Classroom, text books, pop-science books, articles in science magazines, blog posts, half of Quora… At least most of the time, it will regurgitate something semi-sensible, as we are talking about settled fact.

If you say: ChatNazi what do you think of my latest erotic story?

Then the LLM has nothing to go on. The New York Times has yet to review your work. What it does have is generic shit that appears in story reviews - regardless of the actual story. And, as its programmed to fake it when it doesn't have a good match, it just makes stuff up, interspersing a few details from your text with a review of Harry Potter.

And people fall for it.

Yep, it basically acts like someone who took a single creative writing class 20 years ago, but has all the confidence of someone who's been reviewing and writing books for 50 years. I've also had those people give me feedback, and it's surprisingly similar in the complete lack of any useful anything, total misunderstanding, and idiotic suggestions. Those people are great for explaining why they're wrong and in the process of telling them why they're wrong, it shifts your mindset and you figure it out for yourself. But actually relying on it to provide feedback you can trust, use, or take seriously in any respect. Nah.

Every once in a while, it stumbles on something kinda correct. But a bumbling idiot can find the floor every once in a while when they trip and faceplant. So... doesn't mean much.

 

[EmilyMiller]

Every once in a while, it stumbles on something kinda correct. But a bumbling idiot can find the floor every once in a while when they trip and faceplant. So... doesn't mean much.

A stopped clock is right twice a day.

 

[anthrodisiac]

I don’t know if it’s because some of us are ashamed of writing erotica that we fail to reach out to other humans about it.

Some of us have kinks that would make people look at us weird. I already get that enough for being as different as I am, dealt with enough bullying and belittling for a couple lifetimes. Plus, I don't want my existing works and name associated with what I write, because I guarantee you it's not, "Oh, you're the author who wrote X," it's, "Oh, you're the weird pervert who likes animal people." I've tested the waters with people I know, and the look of disgust on their face when it comes us says everything I need to know about how correct my decision is not to come out about it.

Also, kinda shy and have never been comfortable randomly DMing someone about anything, ever, much less to impose upon them for advice about something. Welcome to all the guilt of feeling like a burden most of your life

 

[EmilyMiller]

Some of us have kinks that would make people look at us weird. I already get that enough for being as different as I am, dealt with enough bullying and belittling for a couple lifetimes. Plus, I don't want my existing works and name associated with what I write, because I guarantee you it's not, "Oh, you're the author who wrote X," it's, "Oh, you're the weird pervert who likes animal people." I've tested the waters with people I know, and the look of disgust on their face when it comes us says everything I need to know about how correct my decision is not to come out about it.

Also, kinda shy and have never been comfortable randomly DMing someone about anything, ever, much less to impose upon them for advice about something. Welcome to all the guilt of feeling like a burden most of your life

I’m sorry

But we are all weird in one way or another here. I’m sure you could find your people.

 

[SomaSlave]

A stopped clock is right twice a day.

Unless it's digital.

 

[anthrodisiac]

I’m sorry

But we are all weird in one way or another here. I’m sure you could find your people.

It's okay, I've had a lifetime to figure out how to cope with all this nonsense. I embraced weirdness a long-ass time ago. It's more fun anyway

 

[EmilyMiller]

Unless it's digital.

Nope: unless it’s on a 24 hour setting.

 

[JGittes]

Some of us have kinks that would make people look at us weird. I already get that enough for being as different as I am, dealt with enough bullying and belittling for a couple lifetimes. Plus, I don't want my existing works and name associated with what I write, because I guarantee you it's not, "Oh, you're the author who wrote X," it's, "Oh, you're the weird pervert who likes animal people." I've tested the waters with people I know, and the look of disgust on their face when it comes us says everything I need to know about how correct my decision is not to come out about it.

Also, kinda shy and have never been comfortable randomly DMing someone about anything, ever, much less to impose upon them for advice about something. Welcome to all the guilt of feeling like a burden most of your life

I've taken part in several writing groups and workshops and quickly learned that it's best not to reveal that I write erotica. They may say they attended a signing event with EL James, but to actually admit someone in their group writes erotica for personal reasons is beyond the pale.

This is just my experience, and possibly others have found great writing relationships in such general-admission groups, but that hasn't been my experience. Most participants believe they have NY Times bestseller-list novels or New Yorker short stories emanating from their fingertips. I've even taken excerpts and extensively rewritten them to tone them down to 'adult romance' standards to no avail.

This isn't to say that the workshops weren't worthwhile; they were. I learned valuable writing tips and strategies. What I really wanted, though, was thoughtful feedback on the writing I love to do, and that didn't happen.

 

[TheWritingGroup]

The degree to which so many of my fellow humans are gleefully throwing their brains down the toilet so that they can cede control to a machine is never going to stop bothering me.

I don't mind seeming to contradict myself. I contain multitudes.

You sound like Plato's Socrates, condemning reading because it means people can't use their own memories.