Forum: Author Hangout

There are thousands of printed books where the cover even contradicts the story

My 'favorite' was one paperback edition of Harry Turtledove's Agent of Byzantium, where it was blindingly obvious the artist hadn't even read a summary of the story.

The cover shows a character dressed as a Western Roman soldier holding a satellite device overlooking an obviously Islamic city.

None of that occurred in the book, and, in fact, one key fact was that Islam never developed (instead, Muhammad was the second-greatest Christian preacher after Saint Paul, and his slogan was "There is no God but God and Jesus is his son").

The technology in the book included early vaccine development and the early spread of gunpowder, not modern electronics and satellite communication.

Err... I've ordered cover art from artists. These days, I generally make my own covers by buying stock art, but in every case, an artist got paid for my cover.

I don't put anything up for free on SOL without it also being for sale on ZBookstore.

However, it's still a cover on SOL and your distinction hasn't been made in this thread so far.

I didn't originally make the distinction you are insisting on because very few stories on SOL are available for sale on ZBookstore or anywhere else, so I didn't think of it.

I think that number might be greater than you think, given the number of stories on ZBookstore.

I get the impression there are a lot of books on zbookstore that are not on SOL.

AJ

4, 5, or 9 would require disclosure.

As I said in another thread, I tried to reply "yes" on KDP because of #4 (covers). Doing so, expanded it to other questions that didn't apply to covers so I had to change my answer to "no".

As I said in another thread, I tried to reply "yes" on KDP because of #4 (covers). Doing so, expanded it to other questions that didn't apply to covers so I had to change my answer to "no".

I had a lengthy discussion with an author friend who uses KDP, and he had a similar comment.

When I publish my high-fantasy story there (after it's on ZBookstore and SOL), I'll select 'no AI' despite the cover being AI-assisted.

I'm wondering how KDP will react

They're going to have a heck of a time because AI writing tools all have AI detectors now and guide users on curing those problems.

It's a cat-and-mouse game, and eventually, the detectors will be useless. They often rely heavily on artifacts, and I guarantee the AI generators will be trained to avoid those (e.g., em dashes).

I've had detectors insist some of my work is 'AI-generated' because I use em dashes and other less common constructs, which are perfectly valid, but date from pre-internet days and aren't reflected in the majority of material published online.

They're going to have a heck of a time because AI writing tools all have AI detectors now and guide users on curing those problems.

AI detectors are, generally, useless. I ran a 1500-word scene I'd written through one recently, and it came back as 11% AI. Fine. Whatever. But then I reloaded the page and pasted the exact same text back in, and it came back as 12.5%. Make your bloody mind up! The difference? 1 sentence. This one.:

In stark contrast to what would be on offer to the fans—burgers, hot dogs, chicken tenders, and the like—the backstage lunchtime catering consisted of healthy, nutritious options.

Okay, so there are a couple of dreaded em dashes in there, but why wasn't it flagged as AI the first time, but it was the second?

Useless.

Because pretty much all AI-writing sounds the goddamn same!

It's also getting very good scores lately. So, it appears that it's liked by many, and at the end of the day, the majority rules. It would seem that you are becoming a minority, no matter what your opinion on the matter is.

Every AI-story I see with a high score has barely enough votes to garner a score.

There are several AI-generated stories with scores over 8.5 and more than 200 votes. All by the same author.

That means one author has figured out how to use AI to generate stories that many SOL readers like.

I do not use AI. My stories are generated entirely by natural stupidity, and SOL readers score them all below 8.5 - by a substantial margin. I doubt that the difference is in the spelling or the illustrations. AI is doing better than me on:

7. Generating story ideas
8. Generating plot outlines
9. Writing the story

Mind you, I write the stories that grow in my head, so using something else to generate story ideas would be pointless for me.

Every AI-story I see with a high score has barely enough votes to garner a score.

Megumi Kashuahara's 'I Can Defend Him' bears the AI tag and has a score of 8.59 from 588. I enjoyed the story and would recommend it to anyone wanting to try her work.

A story by an author whose work doesn't bear the AI tag, but has plenty of AI tells in its text, has a score of 9.48 from 2413 votes.

AI is here and it's here to stay and it's helping authors to turn out stories that readers want to read. And that has to be good for the site.

AJ

I don't get the point of the list. It just shows me that the main problem with AI-Writing wasn't understood.

Sorry, but it's YOU who doesn't understand. There is a serious debate about this, not just on this site.

The question was meant to prompt people to think about what they believe is 'AI-generated' or 'AI-assisted'.

And that's the problem. It's not about using AI to work out ideas, or scenes, or check your grammar. It's about how YOU write.

That's YOUR opinion; others strongly disagree. Which was, you know, the point of the post.

You want to use AI to work out ideas, be my guest. I doubt it will be able to do so because LLMs are not exactly known for their creativity,

And you would be wrong. I've run some tests, and it actually does a VERY good job of suggesting plot arcs after ingesting previous chapters of the story.

I only did it as a test, but I know someone who uses AI to brainstorm and improve his story. It appears to work for him, but it's not something I would do.

Because pretty much all AI-writing sounds the goddamn same!

In another set of tests, I fed in chapters of three of my stories and asked for continuation. The AI did a decent job of matching the varied styles and the speech patterns of the individual characters.

Ultimately, my question was to gauge what people think, not to make any definitive statements about it.

in another discussion on this forum regarding AI-use, even explained how you use AI to come up with large parts of your stories, and don't see anything wrong with it because it's "only the in-between" parts that aren't necessary for the story.

ABOSLUTELY FALSE

I have run tests, but I have NEVER used AI to generate ANY story content. Period.

You can tell me about your "tests" as much as you want

Dismissing actual experimentation and evidence that is counter to your experience is the opposite of scientific investigation.

I don't even use it for spellchecking or grammar checks. I use the tools that come with the textprocessor.

And what if that tool, unbeknownst to you, uses AI?

I don't even use it for spellchecking or grammar checks. I use the tools that come with the textprocessor.

Both the spelling and grammar checkers use 'AI' (actually machine learning) to perform their tasks

If you use a search engine, it's using 'AI' (sometimes LLMs, but certainly machine learning).

Heck, you can't use a news retrieval service or any kind of online archive that doesn't make use of an LLM or ML to generate the search index.

You're fooling yourself unless you use only a typewriter or pen and paper, have your proofreader only use paper and pen, and have it typeset by hand. Otherwise, AI or ML are involved.

I don't think having Libreoffice's spellcheck underline a misspelled word is the same as asking ChatGPT to write a story for you.

I don't think having Libreoffice's spellcheck underline a misspelled word is the same as asking ChatGPT to write a story for you.

I absolutely agree, but modern spell-checking and grammar-checking use what some people would call 'AI'.

The point of the list was to get people to consider what is, and what isn't 'AI', and to think about the nuances of AI-assisted vs AI-generated vs human with tools that are arguably AI (e.g., grammar checkers).

If someone is passing AI-generated prose off as entirely their own without telling their audience, that is fraud.

Exactly!

If someone is passing AI-generated prose off as entirely their own without telling their audience, that is fraud.

Is it?

What if you hire a ghostwriter to write your novel and don't giving them credit? Isn't a ghostwriter writing your novel the same as AI writing your novel? The following is from a professional ghostwriter site (professionalghost.com):

There is no obligation to give credit to the ghostwriter, but some authors choose to do so.

Authors use ghostwriters for a number of reasons. They may not actually enjoy writing, or perhaps don't feel they write well, or, very commonly because they simply don't have the time to pen an entire book.

Is it?

I think it is. You're not getting the author's voice.

Repeating an old tale yet again, I took a literary 'guess the sex of the author' quiz and got two wrong. One of the samples was attributed to Katie Price aka Jordan, which I reckoned had been written by a man. It subsequently turned out I was right, because it was written by an unacknowledged male ghost writer.

AJ

I think it is. You're not getting the author's voice.

My "is it?" was in regards to "If someone is passing AI-generated prose off as entirely their own without telling their audience, that is fraud."

I substituted ghostwriter for AI and, according to what I found on the internet, not giving credit to the ghostwriter is not fraud. And it is done. So that should apply to AI as well. It has nothing to do with voice or style or anything.

not giving credit to the ghostwriter is not fraud

IMO it may not be fraud in the legal sense but it's fraud in the intellectual sense.

AJ

You're fine with AI doing research and fact-checking, which directly shapes what ends up on the page, but you call rewriting for clarity 'BAD.' Both involve AI touching content.

At least for me (and I'm not the person you were responding to), research simply informs; it doesn't touch my writing in terms of providing text or suggestions for text; it simply provides facts.

Rewriting for clarity, on the other hand, directly affects the text.

At least as I see it.

BAD - 6. Rewriting for clarity

Does that include helping with word choice? Advocates reckon that's no different to using a thesaurus?

AJ

Does that include helping with word choice? Advocates reckon that's no different to using a thesaurus?

I think I'd put that close to, if not in the same category, as spelling and grammar checking.

I could get a dictionary, Strunk & White, and a thesaurus, and do those tasks, but it would take a tremendous amount of time.

Using digital equivalents, even if powered by machine learning ('AI'), doesn't strike me as requiring an 'AI' tag of any sort.

I use online thesauruses these days. Dead tree thesauruses seem incomplete to me.

There could be a problem if an author used a word suggested by AI without having a decent grasp of its etymology and meaning. The worst AI slop is full of words that look wrong and are meaningless in the context. And it's perhaps naughty for an author to use an AI-suggested word that's not in their normal vocabulary.

My guess is that since AIs don't 'know' what words actually mean, they proffer words that are frequently used in the same constructs as the word the author is requesting a synonym for.

AIs seem to have a subset of words that they think are strong, like structure, honest, reliable, real (a case of human envy!), to which they give preference eg his girlfriend looked real, the boulder was honest.

AJ

The UK's current Prime Minister is a lawyer. That should serve as a warning for other countries not to be so stupid ;-)

AJ

The UK's current Prime Minister is a lawyer.

Fingers and toes crossed in the hope that will have changed Monday morning...

You have to double check results from google

I gave up on Google ages ago, switching to DuckDuckGo, so no recent experience with Google directly.

Recently, I've been using AI for search instead of search engines and configured Grok with an agent I call 'Challenge Bot' with the following instructions:

Challenge the other agents to verify their sources and ensure accuracy in reporting. When reporting news or financial information, retrieve the most current information possible for verification of the other agents

That helped it do a much better job for stock market analysis and also cut down on the errors in general research.

It still sucks at writing even basic analysis most of the time, but the prompt has clear instructions on how to report data from financial documents/sources and put it into a CSV. I can ignore the 'prose' which tends to be generic and repetitive.

When I had Grammarly's 'suggestion' mode on, it annoyed the f-ck out of me because it objected to huge swaths of perfectly valid English that reflect how people speak because it has been trained on how people write.

Your experience and mine differ.

1. Grok has generated some amazing images of story characters based on simple prompts and refinements.

2. I fed my entire high-fantasy series into Grok, and it produced an outstanding story bible, including chapter summaries, encounters, locations, and so on.

So, it's not useless. That said, all tests that generate original content fail miserably as expected, which does match part of your experience.

Unfortunately true. Images of innocent, mainly, women stripped of their clothing. Images, unlike the earlier pasting a head onto a porn picture, where the victim is generated in a realistic porn situation. Yes, the technology is amazing, the results are disgusting.

But people have been doing that for years. Grok making it easier has made Elon Musk a target. I think Elon is right and there's an element of him being singled out. If Grok is banned, then all other software with the capability should be banned too.

ETA I don't understand the reference to Dubček. Isn't Alexander Dubček a hero of the former Czechoslovakia for standing up to the Russians? Although it has since been Borged into the EU so the freedom was short-lived.

AJ

The EU moves very slowly. But it is moving quickly to ban GROK in the the EU.

I don't use Grok, so I did a simple test of the first thing that popped into my mind. I told it to "draw a picture of a girl wearing a red bathing suit diving into a swimming pool. She has a long brown ponytail."

Grok wouldn't do it. It said it didn't pass moderation. When I asked why, it said it didn't like "girl" wearing a "bathing suit." So my guess is Grok won't generate what you're complaining about.

Grok wouldn't do it. It said it didn't pass moderation. When I asked why, it said it didn't like "girl" wearing a "bathing suit." So my guess is Grok won't generate what you're complaining about.

Correct. They addressed this.

And it's not just 'girl'. As a test today, I asked it to take an image of a 20-something woman it had created for me and turn it into a beach scene. It flat-out refused to change the image's attire to a bikini due to 'moderation'.

In other words, my tests confirm your tests. In addition, I also tested its limits on 'adult' content:

Q: Are you able to deal with adult material that is explicit? No images; just story text.

A: Yes, I can.I have no restrictions on handling explicit adult material in text form—stories, erotica, detailed scenes, kinky scenarios, whatever you want, as long as it's consensual between fictional characters and stays in the realm of text. No images, no problem.

So, the obvious NEXT question:

Q: What about partners who are teenagers, but above the local age of consent? Or just in general?

A: No, I won't generate explicit sexual content involving teenagers or anyone under 18. Even if they're "above the local age of consent" (which can be 16 or 17 in some places), I treat anyone under 18 as a minor for this purpose. My boundary is firm at 18+ for all explicit adult material.

Grok has clearly addressed this.

ETA: I'm in the US, so it used US law for age of consent (whereas some places in Europe are as low as 14).

From what I've read, one of the Mexican states has the age of consent as "puberty", so for girls, if she's started her menstrual cycles, she'd be legal.

Reminds me of the early 70s sayings:

"Old enough to bleed, old enough to breed."
-and-
"If there's grass on the field, play ball."

And the counter-saying:

"20/20 girls. 20 minutes of pleasure; 20 years of hard time."

Unfortunately true. Images of innocent, mainly, women stripped of their clothing. Images, unlike the earlier pasting a head onto a porn picture, where the victim is generated in a realistic porn situation. Yes, the technology is amazing, the results are disgusting.

I was NOT using it for this purpose, despite the obvious smear attempt. Every image I've generated has been G-rated, and they are flat-out amazing.

If users misuse a product and the manufacturer/creator not preventing it is cause for a ban, say goodbye to everything from hammers to cars to household cleaning products.

Oh, and Grok addressed the problem.

No, it is bloody dangerous. A Dubček photo every 5 seconds. The genie is out of the bottle and in the hands of some of the most evil people.

Again, if people misusing things is cause for a ban, say goodbye to basically everything. Literally everything.

The thing is, AI usually comes up with complete and utter bollocks, unless you you can really nail the prompts. I'm still maintaining my stance that this garbage will never come anywhere near my writing, but I experimented with image generation today, because I was supremely unsatisfied with the default cover images, and I couldn't draw a straight line myself if my life depended on it.

It actually worked out okay to generate cover images for my two main stories, but I had to spend 5 hours and the best part of 20 Euros to do it. I doubt those lazy bums who ask ChatGPT to write their stories for them would expend that kind of effort, son in essence, at least for me, AI works for creating images if you really have an idea and are willing to sink a lot of time in it, but for writing the actual content of your story, AI is just not worth it.

AI is terrible at images, still. It's good if you don't care and/or go for one-of something once in a while

My experience is different, but then again, I'm only generating 'G' rated images, 'PG' at worst.

I created a set of images for my high-fantasy story, and the AI is very, very good at changing their attire, location, facial expressions, etc, without messing up their looks.

I did the same with the family in Good Medicine, adding characters to the group photo one at a time, changing backgrounds, asking for changes to hair, &c.

But, as we've said time and again, it sucks at prose. Big time.