A Tsunami By Any Other Name

Posted on by Sarah A. Hoyt

Do I HAVE to put up my “AI is going to destroy the novel market” decorations already? I still have my “AI is going to ruin science decorations” up.

Some X-brainiac last week — same brainiac who keeps banging the little drum for “you’re all contributing to the death of culture by not paying enough to artists” (Buyers and sellers, how does that work, even?) — said if a writer used AI for covers, obviously he/she was also using it for novels and “their entire catalogue is suspect.”

Head. Desk. Repeat. See that dent? That’s how many times head went against desk.

This is the sort of argument that’s so wrong it would need to be several miles towards “right” to be merely “Wrong.”

First, notice novel writers using AI for covers. Most of us — with a few notable exceptions, and those people do their own covers — are not artists. Even those like me who took 3 years of art, are not at the level where drawing a cover from scratch is feasible time-wise. I could do it, but it would take me almost as long to do as to write the novel. It’s simply not feasible, from a business point of view.

Second, again always with a few exceptions who are artists and writers and who go extra on the covers, while covers can be works of art, they don’t NEED to be. While people pay attention to covers, they treat it more as a billboard for advertising the novel. The cover in itself doesn’t need to be achingly beautiful or filled with soul searching, or whatever. Sure, if it is, it’s to the good, but if it’s the most beautiful, artistic cover you’ve ever seen and doesn’t signal genre, subgenre and pacing of novel correctly? It fails as a cover.

Would most of us still pay top dollar for a great artist if we could afford it? We definitely would. But the problem is even I — who earn semi-decently — can’t afford that.

Also, the artists I can afford will likely use AI to the same extent I do. Keep in mind every single cover takes at least three and sometimes as many as 10 pieces of render “stitched together” and then having the lighting, etc. fixed, and often the whole is overpainted. it’s just that takes a week, not the year it would take me to paint a cover.

Anyway, does this mean that my novels are “suspicious”. (We’ll leave for later a friend’s opinion that AI is going to overtake the indie market in a tsunami of crap.) Meanwhile let’s discuss my catalogue, most of which has been out for over 10 years.

First let’s get something out of the way: My husband says they advertise AI that can “write novels.” This might be true. The AI might even be able to write novels, who knows. What it doesn’t do, if the state of the art I see in Jane Austen fanfic is accurate, is write DECENT novels. It can’t follow a line of experience and emotion. Someone forgives someone else, then starts looking for an apology AGAIN over and over and over. The best thing about those novels are the reviews.

But let’s suppose some LLM can write a coherent, enjoyable novel? Does this mean I’m going to use it to write my novels?

Why on EARTH would I? I want to write my novels. I mean, I came here to write novels and chew gum, and I’m all out of gum.

No, seriously. The best instance of AI written novels would still need extensive edits. That’s the part I don’t like to do. It would be like when I paid someone to watch my kids so I could clean the house. Completely the wrong way around.

However, let’s suppose AI can write a novel, and a decent passable novel.

So, what is the problem? There are more novels to read, right?

Okay, I can hear the Reeing from here. Please, keep it down to a dull roar.

Objection number one: But the stories will be soulless, and people will be churning them out a million an hour. Right?

Maybe. But if they are soulless, and — let’s face it, Artificial Intelligence is just a Large Language Model — they do tend to be prejudiced towards a certain uniformity and non-originality, then why are you bothered that people will churn them out by the hundreds? There might be some problems with search, mind, but people will rapidly learn to route around AI novels. In JAFF reading the first couple of pages will do. Or the comments. (Oh, the comments.)

Look, there are tons of soulless novels being written by alleged humans. They dot every i and cross every t of the “thing to say” and “the in thing”. They don’t seem to be doing great, honestly, judging by the ever diminishing print runs.

You can’t simultaneously say that AI written novels will be very bad, and that they will steal all our jobs.

Making a lot of money as a writer is never a big chance. But if you miss it, it’s not likely to be because of AI.

Maybe in the future AI will write with the verve and passion of a Dumas and challenging ideas of Heinlein, but — looks meaningfully at midjourney, which is the best of the lot imho — it won’t be the near future. And I suspect it will have as much trouble with emotions as with hands. (For the same reason. Most artists/writers have problems with that area, too.)

Until then, it can write beginner novels, or highly formulaic novels. With a ton of editing. Novelists will be fine. We compete with better every day from Eastern European authors who also seem to churn them out by the dozen.

In fact, what my friend was complaining about with indie being overwhelmed by a tsunami of AI is the same old argument as the tsunami of crap.

Was there a tsunami of crap published. Well, kind of. Except most people realize it’s still a lot of work, even just getting a cover on it and publishing it. But up front there were a million deep-trunk novels published, and they were wretched. (Around 2012 Dan used to do dramatic readings of the worst he found. My favorite was the one that explained what a robot was. And an alien. And…. Like he’d just invented science fiction or something.)

But again, those didn’t overwhelm anything. They didn’t find readers, and sank without a trace. When a book has one review, and it’s one star…. well. Again we’ll be fine.

Is anyone still reeing? Oh, yeah, that guy in the back is reeing about plagiarism. To him I’ll say “be your age.”

Plagiarism has a very exact definition, and LLMs are not engaging in it. (Unlike ivy league college presidents.)

Look, there are a couple of people out there who have bestselling novels taking the general idea of one of mine. In one case, it’s probably coincidence, in the other we shared an agent, and it probably came up in a talk with her, particularly since I refused to do what he wanted with the book.

So, I can sue and take all her money, right?

No. When you hear of J. K. Rowling paying someone 100k because this person also had a character named Harry Potter and had sent her the manuscript, it’s not plagiarism. it’s someone very rich paying a nuisance to go away.

Ideas aren’t copyrightable. Character names aren’t copyrightable. Heck, even plot lines aren’t copyrightable.

Do you know what plagiarism is? Taking large CHUNKS OF TEXT. The LLMs don’t do that, or not in any provable way. I mean, some of the sentences are banal and everyone uses them, but a sentence of two ALSO doesn’t rise to the level of plagiarism.

So no. The LLMs are not committing plagiarism.

They are in fact taking plot lines and ideas, but you can rest easy. Because having seen people run brainstorming experiments, nine out of ten of the ideas they give you are “movie of the week” or a recent tv-series.

You have nothing to worry about.

And by you I mean YOU. You who are actually trying to write what you love in the best way you can. YOU have nothing to worry about.

The precious mavens of churned out politically correct prose? My husband gives me to understand they could be replaced to advantage by a small perl script, no AI needed.

But that’s not your concern. Your concern is your novels.

No artist who is highly original, no writer who is highly original, probably no musician who is highly original needs to worry.

Except with competing with themselves. They do need to worry about that. And it will be a b*tch.

AI? I’m mildly worried about people assuming/running on the idea it is actually sentient. But only mildly.

Yes, they want to entrust medical diagnostics to AI. You know guys, highly distracted, extremely overworked doctor vs. AI. It’s going to be gruesome either way, and sometimes the AI gets the hands right. By a miracle!

AI is in fact a fad. Oh, not the real technology. That’s still buggy as all get out, and will need to have the bugs shaken out. And will. And it is very promising.

But the things/ideas/capabilities people attribute to AI? That’s a fad.

Remember the Radio Flyer? Little red wagon for kids. Nothing Radio about it. But Radio was the new, new thing. So everyone did it. There was also a time when Atomic as the in things and everything was atomic.

AI is like that now. Everything is “AI” even when it isn’t.

So, chill. Let things shake out. Neither Radio nor Atomic finished us off. And this won’t either.

All I want (for Christmas) is for King Harv to put out the Radio Atomic AI coffee! I’ll not only buy a few lbs. I’ll buy mugs and T-shirts with the description!

Now forget AI and go do what you do. And do it well. AI will hold no fears for you.

236 thoughts on “A Tsunami By Any Other Name”

  1. I’ve noticed something recently that I suspect is due to AI. There are many articles on the web where entire paragraphs are repeated several times. I don’t know if this is because someone notices people need to be told something several times to remember it, or if there was a demand for X number of words, and that’s one way of fulfilling it. Either way, it’s irritating. It comes close to making text as slow as video to put an idea across.

    Liked by 1 person

    1. You are absolutely correct. As usual. The area that is getting hardest hit by “AI generated crap” is …. articles and click bait stories.
      TBF most sites edit them minimally. Unless they’re (usually foreign) clickbait.

      Liked by 1 person

      1. I was working as a copywriter a few years ago, just before AI started to become “popular.”

        Even then, it was getting cheaper to generate click-bait Facebook and Instagram content with the tools at the time. I suspect it’s only getting cheaper.

        Like

    2. A lot of financial reporting now has notes that it’s generated by AI, which like a lot of the technical press articles deprives some “writer” of the endless task of copy-pasting corporate press release text under their byline.

      Liked by 2 people

      1. Yes. Which makes me suspect that posts without duplicated text blocks are more likely to be LLM generated. The LLM won’t double-click the mouse (or double-tap the keyboard) on “paste.”

        These duplicates are artifacts of very sloppy, or non-existent, editing, to be be honest. (Well, maybe some are WPDE and its cousins…)

        Liked by 1 person

  2. Coincidentally I was just in the area hit by the 2011 Tokoku tsunami.

    A key thing to note is that the tsunami waters receded pretty quickly. Now those waters did indeed trash the place and once they had receded gas leaks etc. led to fires burning much of the wreckage.

    But the water left, and then humans came along and cleaned up the rest. Now, apart from a bunch extremely large concrete sea walls and a lot of wasteland, there’s no clue that there ever was a tsunami apart from various signs (“water got this far”…)

    Something tells me the AI tsunami will be much the same. It may trash the fortunes of a number of “write by numbers” formula writers, Hallmark movie screenplay writers etc. but even that probably won’t last. It could even end up improving things as readers reject the blander formulaic stuff because they thin it might be AI generated

    Liked by 1 person

  3. I wonder what Boris Vallejo or Julie Bell charges for a book cover? And how long it takes either of them from concept brief to finished product?

    Like

    1. I was chatting several (8) years ago with a gent who at the time did covers for Baen. He started at 40K and went up, and kept some of the rights (didn’t ask which ones). Michael Whelan? Boris? “If you have to ask, you can’t afford it,” I think is the answer.

      Liked by 1 person

      1. I came here to mention Mr. Whelan. I have a collection of his works and he mentioned some of the constraints on book covers (space for text, etc.). Incredible artist, especially when you find out that he painted SOME of those covers with a broken wrist. (The talent, it is in the brainmeats, not the fingers.)

        Liked by 1 person

        1. I love Whelan’s work. I also like the fact that the editors who purchased his works seemed to like the same kind of writing that I do, so for a long time his covers were a good indication that I’d like the book. (Darrell K. Sweet was absolutely useless as an indication. *Everybody* used his covers, so his work graced some gems, a lot of competent works, and a whole lot of rubbish.)

          Liked by 1 person

      3. In the RPG world there was a fantasy heartbreaker around 2000, I own it somewhere but the name escapes me and I’m too lazy to get it, that the authors somehow got a genuine Julie Bell commision for it.

        For my 1000 copy never going to go anywhere RPG that was one hell of an expense.

        And I just looked, Undiscovered from Eilfin Publishing.

        Like

          2. I have it because I set out to collect all of the RPGs Ron Edwards covered in his two “Fantasy Heartbreaker” articles. It was interesting enough but my favorites were Forge: Out of Chaos and Dawnfire.

            Like

      1. I know NMI is NOT Non-Maskable Interrupt in this context, but I can’t recall what title it could.

        Like

          1. Sarah, I think that you meant NML, but you lower-cased the L and that looks like an I for Interrupt in some fonts.

            Like

              1. No, that IS an upper-case “L” if you get right up to the screen. It just looks like an “I” with the ellipsis following it.

                Not a typo – a font problem.

                Like

  4. I’m deeply and profoundly worried about all of the jobs and respect that this can take away from mentally incompetent and fashionable scholars.

    Why, it may become impossible to obtain a position as a mentally ill vagrant ranting on a street corner.

    Tongue is removed from cheek.

    Which is to say that I am not entirely in favor of Ian’s position, but I am hardly entirely opposed to it.

    I think ‘AI’ fashion is potentially a dangerous trap for the management of organizations that are already doing a not entirely perfect job of overseeing programmers.

    Yet, the software technologies that we have labeled ‘AI’ are not profoundly different from the previous state of affairs. It was already true, and has been for decades, that management can use existing software technologies to try to do things that they should not do, and which are unsafe.

    Electronic or electrical control of machines, including heavy machines, has been available since well before the PC in the 1980s. Boiler explosions in the 19th century are arguably partly a result of control loop technology that was almost good enough, but not actually good enough. They were also a key example of a control loop technology that can be implemented in a way that is excessively dangerous if left unsupervised by a human. Which is to say that design hazards are not new.

    Wind and water mills for grinding grain, could be involved in all sorts of safety incidents.

    So these are old problems, but for the correct audience safety conversations can still be productive. I don’t want to say ‘Trust the Plan’, nor that the public should blindly trust experts, nor that people who think that they were trained as experts should rest on their laurels and stop thinking.

    On the other hand, while I identify as incompetent, and as needing more work to bring my skills and practice up to an appropriate level, I may actually angst about this stuff too much. I have chronic gut inflamation, and sometimes I am just physically sick in a way that makes me paranoid about things that are not actually problems.

    Like

    1. Ow. If your meds allow it, you might try lemon/ginger tea. I find it soothing and the taste isn’t bad.

      (I am also noting my gut hotspot calms down when we leave home, which makes me wonder about the local water, even though I run it through a Brita filter).

      Liked by 2 people

      2. I’ve known about ginger for a couple months. Apparently stops some nausea that I didn’t know I was experiencing.

        I decided that today I would have some more ginger tea, and it worked out okay.

        Like

      3. Our local water has an interesting chemistry (the usual lime plus iron-sulfate). For reasons that make no sense, I like the taste. OTOH, if I want the water really cold, I use the Pur pitcher. (Had Brita filters float out of the upper chamber. Ever wall a pitcher? Not completely kidding…)

        Like

    2. Thoughts inspired by your meditation on software, management, and control systems:

      I read somewhere the concept of Moral Crumple zones. You could read it as Human in the (Blame) Loop.

      We do not have social structures to deal with true artificial intelligences – an ordinary person is responsible for setting up an agent system. Normal humans (maybe even the correct ones!) will catch flak if the agent goes wrong – regardless of the level of emergent behavior.

      Like

    3. The way I see it, is that the main threat is to producers of buzzword and jargon laden crap.
      Pseudoacedemics, conmen, and other shady types to be hardest hit.

      Like

      1. Bingo.

        To me, all the handwringing about ‘misinformation’ strikes me as i) dishonest ii) indicating someone with shockingly few filters on their inputs, who is just blindly adopting ideas without testing them.

        Now, to be frank, I can see that a lot of the time I am a pretty careless screw up.

        Like

    4. My son knows a 40 year old programmer who was let go and told we dont need you AI can do your job. I expect said company will be gone soon. Meanwhile the programmer is doing anything because the sentiment is wide spread.

      Liked by 1 person

      1. “don’t need you AI can do your job“

        –

        Been the bane of programmers since at least the 90s and before. Problem is never quite gets there no matter how well things have improved.

        Programmer who has been in the business is probably getting a taste of “you are too old” only now the buss phrase is “AI will do this”.

        “company will be gone soon“

        –

        Depends on what the business is. If the business was a programming business, or had a product dependent on programmers, then probably. If not, well they can always hire an outside company. Just depends on whether they admit it in time.

        “programmer is doing anything”

        –

        Know that feeling. Persistence got me a programming job. One, as I’ve stated before, not only 60% less pay than the last job, but less than I made 1990, 14 years later. Easier when you are the second income But dang I can sympathize.

        Liked by 1 person

  5. Spider Robinson once referred you to a certain SF series as the Perry Rodents. Those series will probably be improved by AI. The rest not so much. But if you’re desperate for fanfic about your favorite characters or want to read a cross over about the Fellowship of the Ring meeting Elric to battle the gods of Chaos…

    Like

    1. …or an A-10 gets warped into Middle-Earth, the pilot helps Bilbo, Thorin & Co. park it in front of Smaug’s lair and they lure him out in front of that GAU-8 30mm cannon… :-P

      Like

      1. David Weber and the author of The Doomfarers of Coramonde did variations on that theme.

        Like

        1. Brian Daley, The Doomfarers Of Coramonde and The Starfollowers Of Coramonde. They found that even an M2 .50 cal machine gun wasn’t very effective against Chaffinch. Running over his head with a 12 ton M113 APC did the job, though.

          Then the only sorceress who could send them home had to be rescued from Hell…

          Like

      2. …or a Valkyrie tank winds up in front of a medieval castle owned by a necromancer, and we find out what megatons-per-second do to the eldritch Dark Ones whose powers we know naught of…

        …or what a demon army does when somebody (whose name is George) releases a Mobile Infantry orbital combat drop on them…

        Coming soon, if I can ever get my cover sorted out. Written and in the can for years now, no cover. This is an example of AI still being too sucky to generate an acceptable thumbnail for a book. I’m photoshopping a couple of pictures from my phone instead.

        While we all wait for me to get my schlitz together, you can read about what happens when humans do wheelies in front of the Space Cops in their shiny new FTL ship that they’re not supposed to have. https://phantomsoapbox.blogspot.com/2025/03/another-new-book-secret-empire.html

        I’d link to Amazon, but it always leaves this huge picture in the middle of the comments section. Rude.

        Of note, on the topic, the cover of Secret Empire is a picture of the moon I took with my phone. The little spaceship is AI generated.

        So I think perhaps the pearl-clutching from the art community is somewhat self-serving, because there’s zero chance any artist was going to make even poverty-pay from The Phantom for a cover. My total take-home for the novel doesn’t pay for the thumb drive I use for backups.

        Like

        1. [Taps foot.] My wallet is ready for the next installment of George and friends (human, robotic, and yes). Please get something…

          I have a “vintage car” calendar that quite obviously used AI with piss-poor prompts to generate the pictures. The hood ornament that looks like a 2′ unicorn horn was bad enough, but the chrome Cthulu/Starship Trooper Bug as the fender mirror took the prize.

          Silly Question Time: Have you talked to the MGC about using AI for a cover?

          Liked by 1 person

          1. Thanks Pete. ~:D

            Work proceeds slowly, due to Real Life taking up my time. (lawnmower repairs, where I am the one making the parts.) I am inspired by your request, I shall redouble my efforts.

            AI for covers, problem being that if you’re not quite good at it, you get an “AI cover”. There is also the cost of the thing, Photoshop is nearly $80/month these days, here in the Demented Dominion. AI programs also seem to be charging quite a bit.

            However there are much more reasonable alternatives such as Adobe Express, that’s what I used last time. It’s just slow, is all.

            Like

        2. WP hates me. Nary a link, but moderation. Again.

          TL;DR (Waiting patiently, for values thereof). Maybe the MGC can give pointers on AI for covers.

          Liked by 1 person

  6. Actually, AI might just be an excellent tool for an author. You may (probably have) already though of ways to make it work for you. Say you’ve got an outline that you aren’t happy with. Use of the AI to generate alternatives may save you a week of cogitating about it. Or you’ve got a buggy paragraph or chapter, and you need it cleaned up or reworked and you don’t have time for a reader to go through it for you. How about writing projects that you’ve done in pieces over the years and the style changes drive you nuts? Maybe AI can regularize the style through the entire work (if that’s what you want.) And here’s the big one in my mind; something that helps you through a writer’s block. Got a story up to chapter 4 and have no idea where to go after that? Have your AI make some projections and suggestions. AI might only be Clippy on steroids, but it’s better for machines to be servants, not mastes.

    Liked by 1 person

    1. Just the ability to feed it a list of non-standard vocabulary words (maybe have it compile a list of ALL vocabulary) or names across all the books in a series and point out possible inconsistencies would be valuable.

      Liked by 1 person

      1. Oooo. Write a series and never have inconsistencies in the stories? Or remove unplanned plot holes? I can think of a lot of movies and show that REALLY needed that, like a bunch of Disney/Marvel products.

        Liked by 1 person

          1. Han Solo voice:  “It’s too big to be a corporate policy. It would have to be a national, or imperial policy.”
            Luke Skywalker voice:  “I have a very bad feeling about this.”

            Liked by 2 people

      2. Making Wikis of your series is in fact a thing AI can do, Dan tells me. And you’re right, that’s a tool I probably will use, particularly with NML where everyone has weird names and over six books it will have a cast of thousands.

        Liked by 1 person

          1. I was starting to consider a spreadsheet for the Familiars books, but decided too many useful parameters/characteristics for the characters was going to make it a slower project; GRRM grade…

            Like

    2. AI as a copy editor, like the robot in Asimov’s story “Galley Slave.” I’d like that.

      Like

    3. I write in a peculiar way and my novel ideas appear fully formed and with force. BUT yes, other writers will figure out how to use AI as a tool.
      I know people who use it as a rubber ducky for plotting, already

      Liked by 1 person

    4. I follow an author named Monalisa Foster on substack, and it sounded like she was dissatisfied with how her attempts to get ai to wikify her world-building by having it read the text.(I want to say her workflow is something like novelcrafter plus either chatgpt or grok). But I think we will get there.

      Liked by 1 person

          2. Dan THINKS there is a specific AI…. appliance? (not the right word) that does Wikis for fiction. He could be wrong. He hasn’t investigated. We were just shooting the breeze about it, and he sees a lot of stuff cross his feed. It’s like my saying “Wait, I read something about a dem scandal involving weasels….”

            Liked by 1 person

        2. Also, one time I read out loud into Word all the paper notes I had built up for the space regency, and had Claude cleanup and reorganize the results. Had to do some additional fixing afterwards (Claude’s training on asteroid classification was not optimal) but it was a lot less painful than trying to type up my world-building notes, or dictate to word and handcorrect the results. But the result was not so much a wiki as a single quick reference document I could pull into scrivener.

          Liked by 1 person

            1. Have you tried anything that runs on whisper ai? I have fairly neutral accent but mumbly diction, and i feel like it understands me better than Word text to speech does. It is a pain to set up and run, so some wheeling of family/friend programmers might be in order. Failing that, it might be an interesting experiment to dictate to Word, take the results to Grok and tell it: this was dictated in English by a Portuguese speaker from REGIONNAME, please keep this in mind while doing (standard transcription cleanup stuff, can look up the prompt I use when I get home if it is desired).

              Like

              2. no. I’ll try to remember it. Dictate to word is HORRIBLE.
                Unfortunately my accent is tampered with by mid range hearing loss, which is why my accent is getting WORSE on its way to incomprehensible even to my nearest and dearest….

                Liked by 1 person

  7. I can confidently say I have no fear of AI taking over writing. It will have no effect on my inability to write anything people would want to read.

    Liked by 1 person

      1. Celia, was that you asking about ALH? I sent an email link to log in, but haven’t seen a response to it. The link’s only good for two weeks. If it wasn’t you, please disregard. Thanks.

        Like

      1. (checks wallet)

        Gulp, I’d give it a shot. You have my e-mail? Just send it to the ALH one on Ace’s sidebar. Thanks.

        Like

          1. maildrop 62 at proton dot me (take out spaces), or leave a post on ALH. Thanks.

            Like

  8. Except with competing with themselves. They do need to worry about that. 

    I’ve seen interviews and articles about both Jethro Tull and Steeleye Span where they’re already saying their biggest competitor for new album sales is their backlists. And this was years ago!

    Like

  9. Not even through the post, but I will note the cover for Jame’s Comey’s future best-seller is dark, muddy and looks like someone swirled a stick through a political of dark greens with a touch of yellow. If publishers don’t use Great Art, what’s this guy bitching about?

    (Beloved will be in the hospital today at least, but says he feels better and is flirting with the nurses. Also wants his laptop so he can do payroll for the restaurant. Am picking up his stuff and took a quick tea/refueling break).

    Liked by 2 people

    1. Glad he’s better! And sometimes WP really does cough up a better word. Alas, most of the time it coughs up a hairball that any self-respecting feline or bovine would disown.

      Like

      1. They’re turning him loose today. With orders not to climb on roofs or cut wood tomorrow. Really. In the discharge orders.

        I feel like crap, which is presumably the effect of stress. Glad to be able to take it easy today.

        Like

  12. “90% of everything is crap.”

    A.I. (AKA Large Data Model processors) use existing data to generate their output.

    THEREFORE: 90% of their input is crap. Crap in, crap out.

    Nobody needed A.I. to generate crappy novels before. LDM’s just churn up the crap and spew it back out.

    “If somebody could just get the LDM’s to differentiate between crap and gold, they could replace novelists!”

    People haven’t managed that trick, and they’ve been trying for centuries. Look at all the crap that’s been written, edited and published without any A.I.

    Heh. Just corrected a typo, but maybe it wasn’t. Good novels are written, bad novels are wrotten? :-D

    Liked by 2 people

        1. “To error is human. To REALLY screw things up requires a computer.”

          –

          “Computer screws it up faster!”

          “Fixing the computer screw up takes longer than doing it correctly.”

          After all no one in my old offices (or any developer/programmer office) never heard the exclamation of “Well … S*!” Or “Noooooooo!” Not once. In 30+ years, not a single time. Maybe thousands but not once.

          Liked by 2 people

    2. Interestingly, there’s a line of research that shows you can get competitive results from smaller models by selecting higher-quality data, rather than just dumping in everything you can get your hands on. I think Microsoft has been looking into it because they want to get their models small enough to run locally on Copilot+ computers.

      Like

  13. My suspicion is that AI will not compete with novels in that way. Rather, as people get comfortable with LLMs and how to use them, you will see people using them to run their own guided stories for their own personal entertainment, rather than picking up a penny dreadful.

    Especially formulaic genres may suffer, because the reader can simply draw a reasonably passable and personally tailored formula out of the box, having a moderately novel experience of a relatively consistent quality. And if sometimes you have to kick it under the table to get it off repeat, that’s just par for the course.

    But I think that is where we will see the real disruption from AI: not in making static books, but enabling live interactive fiction.

    Liked by 1 person

    1. Meh. As a popcorn reader — seriously, whatever my current popcorn genre is, which has gone through space opera, mystery and now JAFF I don’t even REMEMBER them after a few weeks — it doesn’t work that way. I like the variety within bounds, and LLM would be too samey.
      In fact, I stopped reading popcorn mysteries when i realized I’d accidentally read the first half of one, second half of another (different authors and publishers) and it MADE SENSE.

      Liked by 2 people

  14. A LLM, by design, treats language purely syntactically. It can string words together into sentences just fine, and sentences into paragraphs adequately; but it has no idea that the words and sentences mean something. Thus the “hallucinations” of nonexistent legal cases, for example: the LLM knows the syntax for a reference to a court case, but it doesn’t know court cases are actual events out in the real world.

    LLMs that generate pictures have similar problems – there is a “syntax” for pictures, and an LLM can imitate that syntax to generate a picture, but it has no concept that pictures depict things. The problems are less obvious because a single picture is closer to a paragraph of text (in terms of meaning conveyed) than even a short story, but they’re still there.

    Now a story exists above the level of syntax. This is why it’s possible to translate a story from one language to another, and to adapt a novel into a film. So a LLM could not generate a novel or a film, even in principle. The process of thought behind telling a story is something a LLM doesn’t even try to simulate.

    What worries me about the advent of LLMs is the possibility that the basic techniques of constructing sentences, or of drawing and painting pictures, might be lost – that people might forget, not storytelling, but the rules of grammar and prosody; or the methods of drawing perspective, and of indicating light and shape by placement of shading and color. The things a LLM simulates aren’t the essence of storytelling, but they are skills a storyteller must master to convey meaning to an audience. What happens to storytelling if nobody learns those skills?

    Liked by 1 person

    1. Our schools no longer teach spelling, grammar, syntax and sentence structure anyway. High school graduates are unable to read, much less write coherently. They know all about envy, grievance and weird sex fetishes, but their protest signs are misspelled.

      Leftroids managed that without any A.I. at all.

      What A.I. will do is compete with N.S. (Native Stupidity) in churning out crap. I don’t really care who wins that battle.

      Liked by 1 person

    2. What worries me about the advent of LLMs is the possibility that the basic techniques of constructing sentences, or of drawing and painting pictures, might be lost – that people might forget, not storytelling, but the rules of grammar and prosody; or the methods of drawing perspective, and of indicating light and shape by placement of shading and color.

      The stories I’ve heard coming out from schools is that students—who, by the way, are there for a credential, not to learn—are having difficulty functioning without things like ChatGPT. A college professor reported that he’d had students ask for extensions because the internet was down, and found nothing wrong with the concept of not actually writing the papers themselves.

      Seriously, were I in that position, I’d reverse the classes. Everybody, here’s the lecture that was pre-filmed. Watch that for your homework. Come into class, you put your phone in a basket and sit down; here’s your pen and paper. Now write. What, you can’t? You should blame your previous school for not teaching you properly.

      Like

      1. If they’re using the 365 product, all the application is on the internet, not on the laptop or PC; so they literally don’t have the ability to write/type anything if the network is down.

        Liked by 3 people

            1. LibreOffice at least is available now. Back when son was in school there was no other option than the online MS Office Suite provided by the college, that was affordable.

              Like

        3. 365 is now the entire setup– you do need a paid account to use the installed app, but nothing stops you from having LibreOffice and then opening it with the online app to send in to class.

          It’s also extremely common for colleges to provide [desired word processing program] account/subscription.

          Like

        4. I pay for the basic family package through 365 and have the apps on my computer. They work just fine when offline. Yes, I’ve tested it. You just have to go download them rather than assume they’ll be automatically installed. The seismic software company I worked for until 5 years ago did the same thing through the cheap end of the business side.

          Like

        1. Well, to be fair, that is how they are coming out of high school, they are just not doing any improvement during their credentialization.

          Like

          2. To be fair to gifh schools, that’s how they are coming out of elementary and middle schools.

            It’s illiteracy all the way down.

            Like

          3. Not fixing it as effectively as they used to be able to.

            Now, the tertiary schools are partly at fault for what we have available for a lot of the public primary and secondary instructors and administrators.

            Liked by 1 person

      1. It’s the image thing. You look at someone in a picture with their hand on their knee, only three fingers showing, and because of experience you know those fingers are still there, just hidden. AI says oh, sometimes hands only have three fingers.

        You see people in a crowd and see that foot that isn’t attached to anything visible. AI attaches it to the nearest visible human. Oh, now I know that sometimes people have 3 feet.

        Like

        1. Shoot, I got into a big argument with Grok about whether Tip swallowed the wishing pill in The Land of Oz. It insisted it was the Tin Woodman. And that was after I gave it the Gutenberg link. It finally admitted I was right after I copied huge sections of the text into it.

          AI is useful and fun , but don’t trust it for anything important.

          Liked by 1 person

          1. On the flip side, someone I mumble mumble had a professional lawyer give them an utterly incoherent contract, asked DeepResearch for a better one and got something which had only minor easy to fix issues.

            Which matches what I had heard a year or so ago, that competent lawyers are finding that competent AIs are better at creating documents with fewer errors than the [forgets term, younger more apprentice like lawyers] can do.

            Like

            1. Oh, yes, I expect that creating pro forma documents will be one of the things AI excels at. (Still, don’t trust it, and always double check.)

              Like

              1. Exactly; the smaller the possible outputs are, the better it does. That’s why the prompts and training data set are still important, and the weighting of each token by the trainer and user.

                Like

    5. I suspect you could get an LLM to generate a novel (or at least a screenplay) using only tech that we have now, if you really worked at it. Have it generate a synopsis, then an outline, then criteria for individual scenes, then the scenes themselves. At each step of the way, you use a critic loop to have it check for inconsistencies and make sure it stays on script. It’d be cumbersome, but at least in principle, the technical challenge is surmountable.

      Whether to call that telling a story is debatable. It’s still just syntax, with no intention or experience behind it. Even if it works, you’ll see the same kinds of problems as AI-generated articles. But you might get it to produce something that looks like a story.

      Like

        1. Oh, for sure. I just mean that I don’t think there’s a hard barrier on producing novel-length stories, even with the cumbersome workarounds we have to use for limited context windows, inconsistencies, etc. The correct answer to “Can it write a novel?” is “Not the one I’d write.”

          Liked by 1 person

  15. Haha! Love this. Take a chill pill dudes. I remember when Amazon opened up indie publishing. For months it was flooded with “books” with the title of “100,000 DOLLARS!” or variations on that. If you bothered to click on the preview button, you saw a bunch of gibberish or a Serbian phone book. I honestly don’t know whether Amazon cleaned it up or it just faded away under its own stupidity.

    “It would be like when I paid someone to watch my kids so I could clean the house. Completely the wrong way around.”

    Perfect analogy. One that a LLM could never come up with because they don’t produce original stuff. They just take everything they’ve read and put it in a blender. I wrote that story in 1973 when I was a college pup, still wet behind the ears https://frank-hood.com/2025/05/10/writers-cramp/. Never considered it good enough to try to sell.

    For most of us AI might as well be a mysterious wizard’s wand of unknown powers, but some of us, like Dan, come from the biz and know it’s really just the Wizard of Oz, lots of bluster and a little tech razzle-dazzle. That’s why I write about it sometimes to explain it to non techies among us. https://frank-hood.com/2024/01/10/fear-and-envy-of-ai/.

    Thanks for your encouragement to folks out there who despair about being buried by mounds of crap. Do you think it was any easier when your manuscript sat in a slush pile of 3,000 ms. that arrived each month? Agents and editors quickly became jaded and read only the first couple of sentences. Hence the stupidity of writing and rewriting to get the “perfect” first line like, “You’re too late,” sneered Zoltar, “the fireplace has already flown south for the winter,” or “I’m sorry Alex. Even if Father would let me, I could never marry an alligator.”

    Liked by 2 people

    1. I really really really find it very easy to filter out the bad myself. Usually a page and I go “Blah” and move on. If I find a decent writer with a long list, though, I’ll spend months feeding off it.

      Liked by 2 people

      1. That’s why my plan is to build up a long list. I doubt that I’ll write twenty novels, but I hope the few I post and the long list of short stories will be enough.

        Liked by 1 person

      2. I’m almost through Lillian Jackson Braun’s “The Cat Who …” series from the ‘90s.

        Like

          1. I read for the cats and the food. The murders are silly. I like getting library books.

            I grew up with cats and terrible bronchitis. I was sent by my elementary school for tuberculosis screening more than once. Much healthier now without pets, but I miss them so.

            Would you like my print copy of Heinlein’s “Take Back Your Government”? Paperback.

            Liked by 1 person

            1. I grew up with cats, chickens and terrible bronchitis. Turns out no chickens is much better. Still have cats. Might have to buy the food that cuts down on hteir allergen shedding.

              Like

  16. I suspect the main effect AI-written novels will have is to adjust Sturgeon’s Law (or Assertian or Principle or whatever else someone wants to use instead of “Law”).

    Due to AI’s ability to churn out dreck quickly instead of 90% of everything being crap it will nudge up to 91% or 92%.

    Like

  17. In response to the section

    [begin quote] “What it doesn’t do, if the state of the art I see in Jane Austen fanfic is accurate, is write DECENT novels. It can’t follow a line of experience and emotion. Someone forgives someone else, then starts looking for an apology AGAIN over and over and over.” [end quote]

    ….well, that’s just reflecting the actual real-life behavior of SOME people. Possibly the soi-disant “author”. Because a lot of the populace seems to have a very iffy definition of “forgiveness”, and an even sketchier one of “apology”.

    Just my .02, don’t mind me.

    Liked by 2 people

    1. Remember Mr. Collins suggestion of how Mr. Bennett should “forgive” Wickham and Lydia: by stating that they were forgiven, but never speaking to them again or allowing their names to be mentioned in his presence.

      Maybe the AI had figured out certain characters in Jane Austen’s world after all.

      Liked by 2 people

  18. The Turing test is not a test of technology, AI or not. It’s a sociological test, which humans have been failing ever since it was proposed. I’d argue that proposing said test was itself a category error.

    Even the term Artificial Intelligence is values laden with a presumptive humanity that we don’t ascribe to ‘mere’ algorithms.

    So it’s not surprising that public discussions focus on Intelligent Agents Overwhelming Us! – it’s baked into the framing in a way that wouldn’t occur as much with something called, say, a Language Transformation algorithm

    Liked by 3 people

  19. I remember a while ago, some guy put up a zombie apocalypse in space novel. Suddenly (a month or two later) there were about a dozen of them. Nothing new, and the original one wasn’t either. But some folks find that kind of thing entertaining.

    AI will not take the story world by storm. But it could be a bit useful to the writer.

    Liked by 1 person

  20. “It’s just that takes a week, not the year it would take me to paint a cover.”

    Uh. Hmm.

    that feeling when you learn something you hadn’t quite realized…

    Like

      1. I don’t think she’d internalized how long a full on painting for a cover would take.

        Honestly, I still have a heck of a time, from watching folks working now when mock-ups can be done and traded back and forth in a matter of hours, and adjustments made by clicking the “remove layer” button.

        Liked by 1 person

          1. I don’t do painting, but even I know that part of the argument on what is “best” is how fast they dry, and I really like oil paints. (Not enough to manage to go back after it dries ENOUGH, but-)

            Liked by 1 person

        2. For me it’s the opposite. You know how you get faster at the things you practice? I do digital paint/edit things for my work all the time.

          Last summer, I did the portraits for a production of Ruddigore. Took the pictures of the cast in full costume according to how I wanted them, matched them up to the painter I wanted to replicate (the most famous painter from period the costumer told me their outfit came from, with one reversal and the Rodericks being, respectively, Van Gogh and the new King Charles painting that is all red. I mean, I couldn’t pass that up.) Then did the digital overpainting in the correct style.

          Took me 5-6 days for nine portraits. Admittedly, I did manage to give myself a bit of RSI, but I had a deadline to meet because I was traveling out of state.

          This is not as precise as a cover picture would be (lot of shortcuts, no AI though because they get it wrong.) But they still looked good enough from standing distance that I have requests in for two more. And it honestly didn’t occur to me why people were treating me like a miracle worker, because this is what I do.

          Liked by 1 person

      2. Just that it takes me a year to write 60,000 words, but 4-10 hours of work to turn out a completed artwork. Which is why I’m an artist and not a writer.

        Like

        1. Ah. well. My ratio would probably reverse if I spent as much time as art and as little time at writing as I do the vice-versa. We get faster at what we train.
          Though this novel is taking a year. Then again it’s 250k words, and I’ve been sick as a dog.

          Liked by 1 person

          1. You are drawn to words and I am drawn to arts & crafts. It’s kind of a joke that when I’m at summer camp, being an adult who has to stay Out Of The Way, I usually end the week with quite a few craft items, even if it’s only those plastic lanyard things.

            Liked by 1 person

            2. I like it when the camp has specific things for the adults to do.

              Melawaka, high Cascade camp, had leatherwork for adults to make telling the Melawaka story in symbols.

              Where are the troops going summer 2025?

              Liked by 1 person

              1. I am in SUCH big luck. First full week of July is J’s troop is Camp Baker in Florence, Oregon. Last full week of July is G & P’s troop at Camp Meriwether near Tillamook, Oregon. And with Grandma living just north of Eugene, well… Drive all three up, kick out the boys, take J to camp, bring her back, we all spend a week together, kick J out and take the boys to camp. All kids get two full weeks with Grandma and I only have to drive out of state the one time.

                Meriwether has a pretty decent adult program, as I saw two years ago. Don’t know about Baker yet.

                Liked by 1 person

                1. “Meriwether” … We never got there.

                  “Baker”? Do not remember. Then too it has been almost 20 years since I’ve been there. Should still have the walk to the Dune and back for Tenderfoot(?) hike. Plus a 10 mile hike option to the dunes, across the dunes to ocean, and back. Also overnight on the dunes.

                  Which campsite?

                  Grandma’s got to be thrilled!

                  Liked by 1 person

  21. These LLMs aren’t Real AI’s.

    Heinlein’s Mycroft could run circles around them let alone what Skynet could do.

    I’m not worried about these LLMs could do.

    Liked by 2 people

    2. Instead of choosing to take over the world, SkyNet chooses to take over fiction publishing.

      Hmm, might be a story there…

      Liked by 2 people

      2. How would it be any different? Have you seen what’s available in genre fiction these days? *Glares at shelves of regional book store* I suspect AI is used to write some cover copy. Others are so bad, it has to be the author/editor.

        Liked by 2 people

    3. They’re artificial intelligence (the field), not artificial intelligence (the end goal). The problem is they’re flashy, opaque, and really good at faking real intelligence, so people are getting carried away in a way that they didn’t with, say, chess AIs or machine translation.

      Liked by 1 person

      1. Hey, as a retired translator, I can tell you the level of translation you can get from LLMs would seem like black magic to me 30 years ago. …. sometimes still does.
        Not that I mind, I HATE the job of translating. it’s boring.

        Like

        1. There’s kind of a funny treadmill in AI where it stops being “intelligence” when the machine gets good at it. Chess was a big benchmark until it wasn’t, and now we take stuff like translation and image recognition for granted.

          I’ve got some qualms with neural models, but translation honestly seems like one of the unequivocal wins. You feed it text, and it translates it for you. Magic!

          Liked by 1 person

        2. Just don’t ask them to translate a geology article from a relatively rare (Estonian or Latvian) language. I’m glad I could back-engineer the meat of the piece to find the information I was looking for. That gleaned, I just laughed and tried to read the translation aloud. The rest of the group looked at me as if I had grown antlers, or laughed. The interpreter glanced at the original, then at the “translation,” and sighed loudly.

          Liked by 1 person

    4. Yeah, my expectations for A.I. were set by science fiction and I’m not impressed with the current crop of LLM.

      Like

  22. AIs have to be trained. Aside from the legal problems of whose property they are trained on – compare that to training a child. It takes years and exposure to all sorts of other people and different environments to produce a human being worth a damn. Now, imagine you locked a child in the house and only allowed him to learn and form a personality from the internet, with a lot of restrictions on what he could see… Yeah, that’s the sort of AI you will get, too.

    Liked by 1 person

  23. “It’s simply not feasible, from a business point of view.”

    Yes, that.

    Am I really going to pay $200 for a (barely serviceable) cover for an e-book that will, if I am lucky, have dozens of sales?

    It’s a -thumbnail- my friends. Nobody is going to see it full sized until they buy the book. And then they won’t care, they bought it already.

    AI is nice for adding little elements to pictures for those of us who can’t draw, or do so verrrry slowwwwly.

    I note that there is no groundswell of objection from anyone that all the typesetters in the world got put out of business by desktop publishing in the 1990s. This is the same, but for clip-art.

    Also, could we all stop accepting the propaganda that these LLMs are in any way “smart”? They’re as smart as an abacus. They do the same things an abacus does. You could run an LLM on a abacus if you had all the time in the world and a really big sheet of paper to write on. You could run it on a computer made of tinkertoys and marbles. That’s how it is.

    Example, try to get any of the LLMs to draw a watch or a clock, and tell it the time should be any number except ten to two. It will draw ten to two, no matter how you prompt it. Because most of the pictures of watches and clocks it has been trained on are advertisements, and the convention in advertisements is ten to two.

    Show me an LLM that can produce a picture without having been trained on all the art on the entire Internet. Then I’ll think maybe you’ve got something interesting. It still won’t be “intelligent” but it’ll be new, at least.

    Liked by 2 people

    1. I note that there is no groundswell of objection from anyone that all the typesetters in the world got put out of business by desktop publishing in the 1990s. This is the same, but for clip-art.

      There was then.

      Show me an LLM that can produce a picture without having been trained on all the art on the entire Internet. Then I’ll think maybe you’ve got something interesting. It still won’t be “intelligent” but it’ll be new, at least.

      Show me an artist which can produce a picture without learning.

      The definition of “creativity” you have to use to exclude AI results in excluding the overwhelming majority of human output as well. Which to be fair some people do want to define it that way. But that just makes “creativity” irrelevant to most people’s existence.

      Liked by 1 person

          1. Even chimps do art. It isn’t -representative- art, but they do it deliberately and they decide when the painting/drawing/whatever is finished. Looks like kid-art.

            Or Neanderthal art, which certainly is representative, and there was nobody around to tell them how to do it 40,000 years ago.

            One of the foundations of art as a thing is that the person/animal/whatever -knows- what they’re doing. They have an aim in mind, even if they’re doodling in the sand with a stick.

            Ask a 3-4 year old to draw a dog and you get non-representative squiggles and scribbles, because the child has childish motor control and zero skill. But usually you get a circle with four legs and eyes, or a stick-figure dog. Girls generally do the circle because to a girl the face is the important part, boys do the stick figure because the whole dog is important to a boy, not the face. (I raised an artist, I have voluminous evidence of untrained kid-art.) But the child sets out to represent a dog, even if they do it badly.

            LLMs do not know what they’re doing. They have no existence or intent in and of themselves, they are mathematical algorithms performing numerical functions on an input to get an output. You ask an LLM to draw a dog, it reproduces a variation of what it’s database identifies as a dog. (Same thing as the Mandelbrot Set, which is pretty but not art. The Mandelbrot algorithm didn’t set out to be the way it is, it’s arithmetic.)

            These are not the same. The LLM output can be useful in a commercial endeavor if it simulates representational art sufficiently well, but there is nothing of “intelligence” in it. It’s math. The intelligence came from the minds of the programmers and the person who wrote the prompt. The kid made Art. Really crappy art, but still.

            Like

            1. Yeah, I don’t look at a picture to commune with the spirit of the artist anymore than I read a book to psychoanalyze the author. I think doing so is evidence of a diseased mind.

              Intention is only relevant to art in measuring how close they came to their goal. If it’s an AI, then the intention of the person driving it is what matters. But for some reason people want to have it both ways: the AI doesn’t count so it can’t be art because no human intent, but the person using the AI with intent can’t count because they didn’t put in the effort the speaker thinks they should have had to put in to count.

              It’s not about the hunting.

              Liked by 1 person

              1. This is a different argument. You have switched from what the thing -is-, to a discussion of utility, what it is good for. And the nonsensical arguments of people who can’t think clearly, those who want to have their cake and eat it too.

                An LLM copies what humans have done and changes it enough for commercial use, cheaply. That’s not such a bad thing for -me- as an author, because I have a specific need for an inexpensive thumbnail for my ebook. It’s a marketing item.

                Objections are mostly “but what about meeeee?!!!” from all the visual artists. Who are doing exactly the same thing all the typesetters and printers and commercial art people did when desktop publishing started to be a thing in the 1990s. I remember all the whinging I used to hear about it. And again, same thing, desktop publishing was a great thing for -me- because I could make my own flyers, novels, newspaper ads, blogs or whatever, cheaply.

                Fast forward to now, all the printers and advertising professionals use desktop publishing and etc. and they are still working. LLMs will be no different.

                Mostly because commercial art is like a haircut: when you do it yourself, it looks like you did it yourself. During Covid, cutting it yourself is good enough because nobody is going to see you. Going on stage in front of thousands, you go to the good hairstylist.

                So yes, if by some quirk of fate my novels start getting huge notice, I will spend some bread on a -good- artist. Until then, I’m stuck doing it myself because spending money on cover art will not get me noticed, in and of itself.

                Like

  24. we are all going to have to find the thing that we have an A game in to do as a living, C grade and below can now or very soon be replaced with AI. This includes medicine, where AI teledocs can diagnose thru video as well as C grade doctors, AI can code as well as C grade coders, AI can do art as well as C grade artists, and all people that make a living writing 25 things (things to see in New York, things to eat in October, things to make for craft fairs,…..) articles on the internet can be replaced with AI writers.

    I have played with a free, until I want to removed the watermarks, Art AI and have made multiple pieces of art that I would feel were suitable for a book cover in 30min or less.

    Like

    1. we are all going to have to find the thing that we have an A game in to do as a living,

      Which is where the problem will start. Because the left hand side of the bell curve for that is going to be stuffed full of people who either don’t have an A game, or the thing they have an A game in can be done by automation better than their A game.

      However, they will still be here. And what they WILL have, in our warm body democracy, is a VOTE. Or the ability to cause damage.

      And we’ve already seen that humans need something productive to do.

      Like

        1. Sarah, I’ve been working with end users for 40 years….. I never underestimate my fellow humans. 8-)

          Like

          1. Yeah, but if you had worked with me in that situation, you’d think I’m a full-on retard. It’s not so much I can’t tech. It’s that I have to get to it upside down and sideways from any normal techy person.
            Drives Dan BONKERS because he knows I’m not stupid. For the longest time he thought I was faking it.
            BUT I do have the things in my A game.
            My guess is a lot of people can be freed from meaningless drudgery and after some scrambling find their A game.

            Like

      2. That’s also why the utter destruction of the public school system is a sight for sore eyes.

        People tend to be smarter when you don’t lobotomize them as children.

        Liked by 1 person

  25. Websters Unabridged Dictionary contains every word you would need to write the next Great American Novel.

    The skill is arranging them correctly in novel form.

    I have doubts that computer code will rise to that level of skill. It’s not the same thing as being able to compute chess moves 25 possible combinations ahead. Not at all.

    Liked by 1 person

      1. Even Mycroft sentience had its limits. Take jokes: Not funny ever, funny all the time, funny once.

        Like

      2. It requires a person, at the very least. This is one of my areas of interest, and my books are full of talking about it.

        Take your average semi-autonomous combat spider. It will have been given a model for success in whatever it’s doing, be that combat or lippy repartee with humans. It can, with sufficiently clever programming, successfully simulate conversation and jokes in real time. But it does everything based on cost/benefit calculations and a fat database. It’s a computer doing math.

        Whereas a true artificial Intelligence is a person. A person makes their their own model for success, because a -person- is a different sort of thing than a computer doing math. One of the differences is intent. A person HAS intent. A computer is given intent by its programmer, it has no intent of its own.

        To write a novel that isn’t gibberish or a copy of another novel requires intent at the very least. Or so I suppose, anyway. ~:D

        Like

  26. Your title today, “A Tsunami by Any Other Name”, reminded me of something that happened when I was a larval geologist, back in the late 1960’s. At one of the big geophysical conventions, someone got upset because tidal waves were called ‘tidal waves’. Their problem was that the phenomena had nothing to do with tides. I guess they thought that someone, somewhere, might be injured or inconvenienced by the term.

    There was a great search among academics with nothing more important to occupy them which finally came up with an exotic sounding Japanese word- Tsunami. Later, I understand, someone thought to ask what tsunami meant in Japanese. It’s Japanese for Tidal Wave.

    Liked by 1 person

  27. We don’t really know what people will do as their A game.

    For example, a lot of people apparently want to have full-on debates, with spectator chatboxes on multiple channels, about fairly obscure points of theology.

    And apparently some gamer guys have a talent for real-world political analysis, without ever having gotten into politics themselves.

    And people will pay them for this stuff.

    Liked by 1 person

    1. Right? there are things appearing that I didn’t know could exist ten years ago.
      I told Dan that when I was very young I wanted to be be a journalist. I sort of am, for that portion of journalism that involves analyzing things, as well as being sort of the editorial desk for politics at night on Instapundit.
      Neither of these forms existed when I was a kid, and when I was 21 I gave the idea of journalism up because it was so taken over by the left.
      But then…. the Lord really wanted me there and led me by the hand. (Unless you have a better explanation, because I don’t.)

      Like

      1. I’ve had one or two occasions where the Lord decided that “led by the hand” wasn’t sufficient, and He had to choose between nose and grabbing an ear. We may have been made in His image, but the lens was foggy.

        Liked by 1 person

        1. For now we see through a glass, darkly; but then face to face: now I know in part; but then shall I know even as also I am known.

          1 Corinthians 13:12

          And only The Master shall praise us, and only The Master shall blame;And no one shall work for money, and no one shall work for fame,But each for the joy of the working, and each, in his separate star,Shall draw the Thing as he sees It for the God of Things as They are!

          https://www.kiplingsociety.co.uk/poem/poems_earthslast.htm

          Liked by 1 person

  28. “To AI or not to AI, that is the prompt.”

    Recent on YouTube I have been offered “perfect revenge stories” as suggested content. Standard fare is “my wife/husband/job cheated me and here is how I actually knew months/years before and plotted their downfall with sweet justice with the help of my awesome talent or deus ex machina”.

    I sampled a few and determined that they were probably AI generated and upvoted by sockpuppet accounts. Plot holes that you could drive a oil tanker through and repeated details that were obvious as em dashes in a LLM resume or an office cactus plant.

    Liked by 1 person

    1. HOA stupidity video stories is what has been popping up on you of tube whenever I’m on there. They could be true … Regardless stay away from HOAs.

      Like

  29. “Do you know what plagiarism is? Taking large CHUNKS OF TEXT. The LLMs don’t do that, or not in any provable way. I mean, some of the sentences are banal and everyone uses them, but a sentence of two ALSO doesn’t rise to the level of plagiarism.”

    LLMs memorize some percentage of the text they’re trained on, in that they can spit it back out verbatim if you give them the right prompt. Verifying it’s a pain, though, because (1) we don’t know what prompts will cause it to spit out its training data, (2) if it deviates from the training data by a few words, it becomes a lot harder to check, and (3) no one knows what all the proprietary models were trained on.

    One approach is to establish a lower bound by feeding in a passage from the training data and seeing if it continues the passage with no errors for ~50 or so tokens. Depending on model, size, the way it was trained, and how you prompt it, you can get it to spit out verbatim training data ~1-3% of the time. (That’s from a 2023 paper. Not sure whether things have gotten better or worse since then.)

    This doesn’t affect your larger point, though, and in practice, you’re much more likely to get something stitched together rather than something directly plagiarized.

    Liked by 1 person

  30. Faddishly naming things is stupid, limiting and displays a dearth of intelligence.

    My favorite? Robotic surgery. There’s nothing robotic about it. The only intelligence involved is the surgeon himself and the ‘robot’ has exactly zero independent, or even programmed action. The surgeon actively controls every second of the process. Science fictionally, this would be called ‘waldos’.

    Like

  31. Stirling explicitly stated the 1632 books and his Nantucket series (“Island in the Sea of Time”), and the 1632 books and his Emberverse (“Dies the Fire”) series, couldn’t be considered plagiarized, even though similar concepts behind the books and the series.

    Liked by 1 person

    1. Suppose to be sub reply to Stan’s post above. But works for the AI plagiarize general post too.

      Like

  32. There is one thing that AI will never replace, via America’s Newspaper of Record of course:

    Like

  33. Earlier this year, out of curiosity, I asked ChatGPT for a review of Mackey Chandler’s April series. It gave me a review I thought plausible—but of course I did: I recognized many of its statements as taken from my two online reviews of those books (one at Black Gate for the first volume and one at Prometheus for the whole series). I pointed out that what it had given me was strikingly similar to published reviews I was familiar with, and it said that when there were fairly few reviews of a work, its sampling methods would replicate them more closely than usual. So there are cases where ChatGPT does something very much like plagiarism. (I would also say that not enough people are reviewing those books! They deserve more attention.)

    When there’s a lot more material to use as a basis for sentence completion, I’m not so concerned; the “theft” from any one document or any one personal is statistically tiny, and you probably could not find a meaning correlation between its output and any one source. And, well, when I write, MY output is somewhat based on all the things I’ve read, too.

    On the other hand, that very fact determines the limits of what LLMs can do. All their output is based on the statistical probabilities of a large corpus of existing material; so it can only come up with things that have a nonzero probability in that corpus. That works against originality. I can get things out of chatbots that surprise me by providing the right sort of prompts, but if I’m going to put that much imagination into it, I might as well figure out what I want on the page and write it myself.

    And then there are all the times the chatbot gets befuddled and loses the names of people who appeared earlier in a discussion. . . .

    Like

    1. “their output is based on the statistical probabilities of a large corpus of existing material”

      I am suddenly convinced that the most accurate use of LLMs is to win at Family Feud.

      Like

  35. At least the author using AI can generate a cover knowing what she wants to show. In the 60’s the artists never seemed to have read the book or even the blurb on the back cover.

    Liked by 1 person

    1. Or at any time, really. I was blessed in that the Baen artists for DST series asked, but…. some of the earlier covers like the first hard cover of Draw One in the Dark…. SHUDDER.

      Like

    2. I think some of the 1960s-70s cover art was done under the influence. Influence of what I leave to the imagination (or experience) of the reader.

      Like

    3. Apparently some art directors back then refused to let cover artists read even a summary of a book’s plot, much less actually reading the manuscript. They would just say “I want a blonde in skimpy clothes with a gun,” and that was it.

      Like

  36. My husband gives me to understand they could be replaced to advantage by a small perl script, no AI needed.

    Could you ask Dan his golf score on that script? I’ve got an office full of Perl coders who’d give a run at beating it.

    Like

  37. I used to be a voracious reader of science fiction. I can get through 2 novels in a weekend. However, the quality of what is available has declined to the point, where I don’t even bother looking at the store anymore. As as result of this, I haven’t read a work a fiction since 2021, nor am I currently inclined to put an end to the non-reading stretch. But lets say that the AI authors don’t make this 1000x worse – for what it’s worth, a clearly AI-art cover would get skipped in my scan. I judge books by their covers, 100%, when shopping.

    Like

    1. What exactly do you think you are skipping over by avoiding AI covers?

      Ignoring for the moment how many people identify anything which isn’t actively hideous as being AI, which to be fair is something the artists taught them to do.

      Liked by 1 person

        1. They deserve the resulting dearth of novels they are left with.

          Maybe they can commiserate with the people who need a protag which is an expy of themselves.

          Like

    2. What the actual heck?
      WHY ARE YOU LOOKING IN THE STORES? Look online.
      I recommend you go find Space Station Noir now.
      Also, I roll to disbelieve you could tell an AI cover.

      Like

  38. Argh. The following outline more or less showed up for me.

    A Juneteenth Carol

    George the democrat is alone and in that space between waking and sleep.

    Then the ghost of James Earl C., his old business partner, shows up in chains from hell, purgatory, or that big non-dominational and theologically ambigious unitarian chain gang afterlife. Warned to expect three spirits.

    First comes the spirit of Juneteenth Past.

    Which is maybe a long historical dive, John Brown through Reconstruction. My feeling is that the spirit of Juneteenth past has difficulties communicating, and maybe becomes a bit frustrated with George’s intentions to take away a simple vengence oriented narrative.

    Having the spirit be Komi, well, I am not familiar enough with the source material for that.

    Musk would be an amusing troll for the spirit of Juneteenth present.

    Issue is, he is not the correct messenger for the nuance I have in mind. IE, budget cuts as a least bad remedy for mitigating the concerns of the consensus about courts, universities, and the feds.

    Last would be the spirit of Juneteenth future. Again, I feel that George walks away with a message not intended.

    George is a bad person, and a get along, so his post awakening behavior is to still do bad things, but he turns his coat, and instead tries to do all the usual Democrat things to the Democrats who were slower to turn their coats.

    Like

  40. As y’all know i have extensive 3d/CGI experience

    Right as AI became viable, i was considering doing covers for people. technically, i started fiddling with it right before…

    the problem is, literally i would have needed to charge $300-$500 a cover just to cover my time, and even before AI the market wouldnt bear that

    Liked by 1 person

    1. My books usually make more than that even right on first publication, indie. (Over time they make what I used to get as an advance, on average.) BUT my collections of pre-published short stories usually only make about $500 the first year. Paying that for a cover would be inane. And there’s the “I never know if I’ll make a few thousand or it’s a dud….”

      Like

      1. yes, and i understood that at the time, but i literally would have had to ‘replace other work’ in order to have the time to do covers. i made a couple test covers and was looking at how much time it took, and the answer was using largely premade assets at least two nights’ work. that’s why i kinda silently stopped talking about that

        Liked by 1 person

Comments are closed.