back to article Sci-fi author 'writes' 97 AI-generated tales in nine months

Sci-fi author Tim Boucher has produced more than 90 stories in nine months, using ChatGPT and the Claude AI assistant. Boucher, an ML-using artist and writer, claims to have made nearly $2,000 selling 574 copies of the 97 works. Each book in his "AI Lore" series is between 2,000 to 5,000 words long - closer to an essay than a …

  1. Simon Harris

    “Impressive returns”, or not.

    Running the numbers, if I’ve understood it correctly:

    97 ‘books’, each of which takes 6-8 hours to write (call it 7) means 679 hours of work.

    $2000 income from that means each hour’s earnings is about $2.95.

    At that productivity rate I might question whether it’s actually worth it!

    1. Arthur the cat Silver badge

      Re: “Impressive returns”, or not.

      You'd get better pay working behind a bar and interact with fellow humans.

      1. Anonymous Coward
        Anonymous Coward

        Re: “Impressive returns”, or not.

        As an author of science fiction novels, the only thing AI does is muddy the waters. It takes solid characters, a sensible story and proper character representation to make a decent sci-fi story. Anyone doing that in a AI ChatBot will get decent output, but there are no shortcuts.

        G.I.G.O. is as applicable to AI as any other part of life.

        BTW, those seeking to make $$$ from AI need to realize the output is for non-profit use only. Read the Terms and Conditions before you put it your ebook up for sale on Amazon. I suspect the first AI book that sells millions will be sued for any proceeds.

        1. Wellyboot Silver badge

          Re: “Impressive returns”, or not.

          >>>sued for any proceeds<<<

          In a hearbeat if it features space wizards with glowing sticks

      2. Anonymous Coward
        Anonymous Coward

        Re: “Impressive returns”, or not.

        Not interacting with fellow humans is what led me to a career in IT in the first place.

    2. Anonymous Coward
      Anonymous Coward

      Re: each hour’s earnings is about $2.95.

      1. it's just a start

      2. if you 'do' it in your spare time (see 'the argument clinic), you're still 2K up. Like those (...) mining bitcoin, 'every little helps' (and f... the wider cost, cause it's not me who's paying, etc.) People will keep fighting to be the one on top of a sinking lifeboat, rather than plugging holes together.

      1. Peter2 Silver badge

        Re: each hour’s earnings is about $2.95.

        At the point that you have more than 5 articles or short stories published then i'm given to understand that you are considered to be a professional author. If he's published a hundred then he'd be considered a professional author.

        I know quite a few authors ranging from indie self publishing to being acquainted with several reasonably big names in their genre. (as a point of fact, I have an unpublished novel from one of them at the moment for proof reading)

        Several of them make a fair bit of money from talks and speeches at various events at weekends to get some extra money in during the event season for the extra money. I'm fairly sure that even the indie self publishing chap would consider sales of $2000 in 9 months of a new book (let alone all of his books!) to be a life altering catastrophe which would probably force him out of business and into paid employment to make ends meet; I suspect that such a disaster would be a source of deep depression, and something which he'd never mention in the company of other authors. It would certainly not be something of which he'd be saying "look at me!" in relation to.

        1. Simon Harris

          Re: each hour’s earnings is about $2.95.

          You have a good point - I was trying to find out about the economics of self publishing, and came across an article suggesting many authors make more from talking about what they’re writing about than they do actually writing about it, and the books are more a way-in to doing the talks.

          1. Peter2 Silver badge

            Re: each hour’s earnings is about $2.95.

            You have a good point - I was trying to find out about the economics of self publishing, and came across an article suggesting many authors make more from talking about what they’re writing about than they do actually writing about it, and the books are more a way-in to doing the talks.

            Most authors do an absolutely huge amount of research on obscure topics for their writing, and so have something interesting to say at talks. Someone who's told an AI "write me a story" probably has not done that research and probably has little of worth to say.

            In terms of economics, if you make £5 per person giving a talk and then multiply by an audience of ~250 people then you'd end up walking away with £1250 for an hours work. If you do the same talk on a Saturday and Sunday at a two day event then you'd walk away with £2500, which is greater than the yearly total of the sales of the chap in the article, but that's mostly down to him not really being much of an author.

            If your self publishing then you also offer a discount on a signed copy of your book series at the end of your talk and make a bit more cash from doing that.

      2. Michael Wojcik Silver badge

        Re: each hour’s earnings is about $2.95.

        if you 'do' it in your spare time

        ... you're still paying opportunity costs, and the cost of your own labor, and any capital that has to be invested.

        If you're doing it for intangible reasons – because it's a hobby, or an experiment, or whatever – that's fine. But if you're doing it for the income, then the "every little bit helps" argument only applies to the extent that the return on investment is large relative to alternatives.

        If you're constrained in what types of work you can do, then that decreases the opportunity cost and so increases the rate of return. If you're constrained in what you want to do, that increases the return (because there's an intangible psychological benefit). But merely saying "every little bit helps" is sophomoric; it's not a rational calculation.

    3. IGotOut Silver badge

      Re: “Impressive returns”, or not.

      The difference is, just like printed design, it's a create once, earn repeatedly.

      So that income could repeat every year for the next 50, it could fall, or could massively increase.

      1. werdsmith Silver badge

        Re: “Impressive returns”, or not.

        Or a large media company could buy the screen rights to a title or two.

        1. Michael Wojcik Silver badge

          Re: “Impressive returns”, or not.

          That'd be a poor investment, since they could produce something equivalent but sufficiently different to avoid IP encumbrance by the same process.

    4. Michael Wojcik Silver badge

      Re: “Impressive returns”, or not.

      Nor does the "impressive returns" claim in any way support the thesis it's attached to, which is that 2000-5000 words constitutes a "book".

      I haven't seen any of these, and with the pictures they might well have a reasonable claim to be graphic novels. Nor do I care particularly whether they are or not. But if he's going to make an argument, he could at least make a coherent one.

      1. John Brown (no body) Silver badge

        Re: “Impressive returns”, or not.

        Each story can be sold as a "book" purely because they are e-books. The economics would never stack up in the dead tree world so you'd have to stick a bunch together and call it a collection or anthology. There are plenty of "classic" shorts out there of those word counts, even some very much shorter, but would never be published standalone. It's how many aither get started, being published in magazines etc. Surely you bought Playboy to read the articles and stories like the rest of us. Or did you just buy yours for the pictures? :-)

  2. Blackjack Silver badge

    [Each book in his "AI Lore" series is between 2,000 to 5,000 words long.]

    I know people that can write that much in a day, and is honesty not impressive to only write that much with A.I assistance.

    1. Michael Wojcik Silver badge

      Well, rate of production for fiction prose varies wildly, even among authors who can be considered "good" by some metric. Anthony Trollope famously wrote in his "spare time" (after working his regular job of census-taker), according to a strict schedule he devised, and cranked out his novels – which are widely considered at lest good entertainment, and in some quarters serious literature – quite quickly. Kazuo Ishiguro has an anecdote about spending one morning adding a comma to one of his manuscripts, and the afternoon removing it.

      But, yes, in general there's nothing particularly unusual about being able to generate a few thousand words of fiction prose1 in a day or so, particularly when the author is in the groove and fleshing out a structure that's already well-developed.

      1Or substantial non-fiction, for that matter. When I was in graduate school it wasn't unusual for me to produce 5000-word essays in a day. That represented quite a few hours of research, sketching out arguments, outlining, etc over the previous several weeks; but the actual writing I typically did in a few passes within a 24-hour period.

  3. David Pearce

    There are many "real authors" who have done nothing more than rehash existing stories without an original idea over the years.

    This is what the AI basically does

    1. Charlie Clark Silver badge

      Indeed, genre fiction is very formulaic. Many authors will own up to this and milliions of readers of romantic, crime, science fiction are more than happy with it. Same goes for half the production on TV.

      1. Michael Wojcik Silver badge

        Some genre fiction is very formulaic. Much of it is innovative. Generalizations like yours are unfounded and insulting.

    2. Richard Cranium

      William Wallace Cook was doing this a century ago which led to Plotto, essentially an algorithm for creating plot outlines. I suspect it is/was well known to some of the prolific authors of low grade fiction since the 1920s.

    3. Ancientbr IT

      I think you're confusing machine learning (ML) with artificial intelligence (AI). The latter can extrapolate/interpolate, whereas the former cannot. Both are based on neural networks, but their purpose is different.

      An AI can explore the space in and around its training set, based on its nature as an almost-but-not-quite perfectly fitted polynomial curve, extracting novel information.

      With ML, the same input produces the same (trained) output (which is what you want). With AI, the same input can yield multiple, different outputs (not to be confused with "hallucinating").

      With ML, if you input an item from the training set, you will get an output that corresponds to the trained (and expected) result (if it has been adequately trained). With AI, it is incredibly hard (but not completely impossible) to reconstruct an item from the trained set, given the correct specific input.

      For some reason, the popular view is that an AI is nothing more than a database and all it does is regurgitate items from that database. A moment's thought will reveal that that cannot be true, otherwise everyone who entered exactly the same prompt would get back exactly the same response - and they don't (unless they're asking for factual information, but even then, the presentation will almost certainly be different each time).

      In addition, the volume of data on which recent LLMs have been trained runs into the hundreds of gigabytes and more, whereas the size of the trained model is measured in tens of gigabytes. HTH.

  4. Anonymous Coward
    Anonymous Coward

    Did he correctly credit the sources

    in the dataset that the AI trained on?

    1. Spazturtle Silver badge

      Re: Did he correctly credit the sources

      Do human authors credit every book they have previously read?

      1. GruntyMcPugh

        Re: Did he correctly credit the sources

        Indeed, this has been discussed recently at El Reg, with music labels trying to claim copyright infringement for AIs ingesting music, but to me it's the same as an artist saying they were inspired by The Beatles or Hendrix, everything is influenced by what came before, so why is AI suddenly in the cross hairs?

        1. katrinab Silver badge
          Megaphone

          Re: Did he correctly credit the sources

          Because the training data is the source code.

          If you were to obtain the source code for Microsoft Office and recompile it, the resulting binary might look different to the ones Microsoft publishes, especially if you were to compile it for a platform/CPU architecture that MS doesn't currently target, but you would still be infringing their copyright.

        2. breakfast Silver badge
          Mushroom

          Re: Did he correctly credit the sources

          But it's not the same is it? Because there isn't an artist here, there's just a machine making averages. There is no spark of creativity, nothing original, no person involved, only the history of other text, almost all of it used without permission.

          The difference between an artist being influenced by other art and souped-up autocorrect being able to predict entire paragraphs or musical phrases based on the output of other creators writing, is a real apples to statistics comparison and people not being able to grasp that it exists is a profound condemnation of how we have collectively failed to teach and learn critical thinking.

          1. GruntyMcPugh

            Re: Did he correctly credit the sources

            It's not a _lack_ of critical thinking, rather it's being critical of humans, and us thinking we're oh so clever and doing something special that AI cannot.

            1. Michael Wojcik Silver badge

              Re: Did he correctly credit the sources

              There's a large gap between "doing ... something that AI cannot" and "doing something that LLMs don't currently do". You don't have to be a dualist to believe that we're not going to get any particularly impressive literature out of GPT-4.

          2. Michael Wojcik Silver badge

            Re: Did he correctly credit the sources

            This would be a more persuasive argument if you could define "spark of creativity".

            Per my previous posts, I think the current crop of LLMs, and the simplistic unidirectional deep-layered transformer architecture in general, are highly overrated. I don't think they do a good job at simulating the processes humans use to produce what we call "creative" text. But I also don't think that handwaving gestures toward undefined mystical essences clarify the problem in any way.

            (What might be a better argument? I'd start with the Multi-Agent Model hypothesis; point to the heterogeneity of those agents, particularly the abundance of human senses (including e.g. the "passage of time" sense) and qualia; emphasize the importance of unconscious and somatic (see e.g. the research of the Damasios and their team) influences; consider a framework such as Pickering's "mangle of practice" for analyzing how inputs have unexpected effects on human cognition; suggest a high-level arbitration process – think "concept rectification" – based on attention economy; and note the important role of distraction. It's also useful to have a more-sophisticated model of how human languages work in practice – I like Davidson/Rorty – and how the writing process works.)

        3. ChoHag Silver badge

          Re: Did he correctly credit the sources

          For the same reason we distinguish between stuff that's natural and man-made even though all that we produce is no less natural than a beaver's dam from a particular perspective, but it's not a perspective that's particularly helpful.

          Is a skyscraper natural, or the plastic island in the pacific? Does an airport just grow on its own? AI regurgitations feel the same. As related to human creativity as a skyscraper is to a bowerbird's nest.

  5. Anonymous Coward
    Anonymous Coward

    motivation

    According to an interview he says he started doing it to help deal with "mild PTSD" caused by being a content moderator

    https://www.prnewswire.com/news-releases/triumph-over-toxicity-how-a-former-content-moderator-is-reinventing-sci-fi-with-ai-lore-books-301829195.html

  6. sabroni Silver badge
    WTF?

    "more ways to interact with the chatbot so users can search for information"

    Where information means "strings of words that can be parsed by humans".

    What value can you place on that "information" if you have no idea whether it's truth or fiction?

  7. Andy 73 Silver badge

    We are surrounded by morons..

    A very early adopter who gets into this technology before the rush therefore has relatively exclusive access to a tool that anyone will soon be able to use.

    So the fact that he's managed to scrape just $2000 from 9 months of immense 'productivity' tells us all we need to know about the chance that ChatGPT is going to disrupt the creative industries.

    It's like suggesting the people who produce supermarket muzak are going to take Taylor Swift's job.

    (That said, it's notable how badly the current scripts working their way through film and television companies are being received - humans don't seem to be doing much better at producing meaningful work right now)

    1. GruntyMcPugh

      Re: We are surrounded by morons..

      I just asked chatGPT to "create a new literary character, unlike any that have bee featured in previous books, and tell me their origin story' and it gave me a female Dr Who type, who lived in a forest and had magic powers, so hardly unique. Still, 'hardly unique' typifies a lot of action movie screenplays.

      1. doublelayer Silver badge

        Re: We are surrounded by morons..

        Fine, but can you come up with a character that's actually unique? If you're limited to three attributes, you basically can't; there are too many books. Of course, if you let the program keep going, you would eventually get uniqueness by adding more and more attributes to the character, increasing the possibility space until the chance of having that exact character are lower. The chances are that the program wouldn't generate an interesting character, but uniqueness is easier.

        Asking for a character that has literally never been seen before will inevitably get into arguments. Someone who wants to argue that the attempt succeeded will point to tiny details and say that, because no character like yours discovered at the age of five that she could converse with frogs and only frogs, that the character is now unique. Someone who wants to argue that the attempt failed can ignore most aspects of a character and say that, since the character was also trained in the use of her magical powers by an annoying older magical guy who refused to answer perfectly normal questions, that it's formulaic no matter what other things the character has. Character uniqueness is too subjective to be proven and not really the point of the author's creative effort, which is and should be more focused on making an interesting character.

        1. GruntyMcPugh

          Re: We are surrounded by morons..

          No I can't, and chatGPT can't, but there have been arguments made here that AI lacks 'creative spark' and is just regurgitating training material. I don't see it as being that different from an artist being influenced by the works of another and, having shared accommodation with several guitarists in my time 'creative spark' sounds a lot like pointless noodling, until eventually something that might be something is randomly come across.

    2. Rob

      Re: We are surrounded by morons..

      I wouldn't be surprised if ChatGPT is the new 'Bitcoin', a fad that rumbles on for ages that gets over hyped and takes a long time for people to realise it isn't what they expected it to be. Borderline solution looking for a problem and no-one has looked at the mid to long term future of it that ends up keeping it niche.

      1. katrinab Silver badge
        Meh

        Re: We are surrounded by morons..

        If you give it a bunch of facts and ask it to write them up in a particular style, it seems to do a good job of it. So in spaces currently occupied by the likes of Gramarly and DuoLinguo, I think it could be useful.

      2. vtcodger Silver badge

        Re: We are surrounded by morons..

        You're likely right that AI is the next big fad. The trouble is that unlike bitcoin/blockchain, AI has at least one obvious application. It can and almost certainly will be used by large corporations to cheaply provide a new low in "customer service". What do you think your chances are of convincing a chatbot that you did not order 27 large anchovy and pineapple pizzas at 8:30 am on July 16th and the charge for same should be rescinded?

  8. Apocalypto

    AI drafting job descriptions and CVs.

    Warehouse packing robots dying from exhaustion.

    They're really trying to replace human workers in every aspect huh

    1. breakfast Silver badge
      Unhappy

      From what I can tell they're mostly trying to free us from creating art, literature, cinema and music so they can free us up to work at Starbucks for minimum wage.

      Thanks Lads. Best possible use of your time there.

      1. werdsmith Silver badge

        I've considered this and many hours of free leisure time to do what you want with your life feels like it should be the ultimate aim.

        1. Michael Wojcik Silver badge

          It can be argued that under that sort of surplus economy many/most humans would find productive tasks to undertake for their own sake. That's basically the milieu of Doctorow's Down and Out in the Magic Kingdom, for example, where people form "adhocracies" to work on projects that inspire them. Some would even argue that there's historical precedence in some pre-modern societies; you can find examples in the wide array of sociopolitical systems Graeber & Wengrow document in The Dawn of Everything.

          Personally, I find my life most rewarding when it includes a substantial amount of productive labor – I like to get things done – and I like to have some of the requirements for that labor come from external sources, as puzzles or problems that I'd be unlikely to conceive of myself, and so that there are external stakeholders to motivate me. So I wouldn't say that what I want is to maximize "free leisure time". But I would enjoy being able to work on interesting projects in an environment of abundance.

          1. breakfast Silver badge

            I like to get things done too, but if I wasn't tied to having to earn enough money to be able to live I'd be creating music, art and literature, which is much of how I spend my free time. Feeling like the AI engineers are determined to make finding any kind of audience for those things harder is deeply frustrating. Honestly it puts me off their whole field to a degree, which is a shame as I'm sure it's useful for a bunch of stuff.

        2. Ancientbr IT

          Agreed - but with the proviso that a Universal Basic Income is implemented so that core needs are taken care of.

  9. Howard Sway Silver badge

    Write 10 books a month

    Earn 10 bucks a month. That'll be the level he soon arrives at.

    As a science fiction "author", he's surprisingly not yet seen that some kind of random generator of tropes, character types and plots could generate the input to feed into the LLM in huge volumes too, producing hundreds of these crappy "books" per hour, which could ultimately make him several dollars richer, and the world a little bit worse.

    1. doublelayer Silver badge

      Re: Write 10 books a month

      Only if he can find a way to distribute that many books. At some point, his cranking up the supply of crap books will just inoculate people who don't want to see any more of that, and they will go to more lengths to ignore them. If that means that, by spamming out all of these short stories dressed up as books, he ends up decreasing the demand for any self-publishing platforms that continue to host them, that's too bad for those platforms.

  10. breakfast Silver badge
    Thumb Down

    Having just written most of a book the old-fashioned way, it's depressing to know that when I publish it no matter how good it is, it will be no more than a tiny blip of signal in the fast sea of noise from projects like this. It was hard enough for people to find your art already, this is just going to make it more of a struggle.

    That doesn't matter too much for me - I have a day job that pays the rent, but for a brilliant kid writing their first book and hoping they can make a living off it? This is going to make things just a little bit worse.

    1. Anonymous Coward
      Anonymous Coward

      Self publish hell

      TBH self publishing via Amazon (other book sellers are also available) already has managed to damage the book writing industry in a way its unlikely to come back from.

      Every man/woman/dog is writing "books" now, using self publishing and the place is awash with garbage. Combine this with all of the review stuffing that goes on in that self publishing world and its largely game over.

      People selling "books" (or even using them via Kindle unlimited) has mangled the real publishers because readers are unable to see the wood for the trees now. All they see is a "book" that costs 99p vs another actual book that costs £9.99. Guess what happens when the publishers can no longer sell fiction in any great quantity?

      I don't think ChatGPT is going to make this much worse than it already is, which is dire.

      Everyone gets to make $0.00001 a month and all of the quality is crushed - much like so many other creative industries have been.

      (I'm sure I'm about to get slammed by some of those "amazing" self published writers out there now)

      1. Anonymous Coward
        Anonymous Coward

        Re: some of those "amazing" self published writers out there now

        Self-publishing allows fantastic authors to create incredible gems such as this (https://www.amazon.com.br/FALL-MOON-ASTRONOMICAL-OBSERVATIONS-English-ebook/dp/B077YYBY1Z) -- which is *not* science fiction.

        I got a free copy since I work in a scientific institution and the author thought I could do something about the moon slamming on the Earth.

        Great reading for when you're out of booze.

      2. katrinab Silver badge
        Meh

        Re: Self publish hell

        Why would you pay 99p for someone else's ChatGPT output when you can ask it to write one yourself?

      3. Michael Wojcik Silver badge

        Re: Self publish hell

        To be fair, even before ebook self-publishing pulled down the capital costs of vanity publication, we were already producing books at such a furious rate that only a small portion could get widespread attention. And there was certainly no correlation between popularity and quality, by any metric (other than the tautological one of popularity). I can name a dozen contemporary young-adult fantasy novels off the top of my head that are vastly superior to the Harry Potter series in every way (logical and intricate plotting, well-conceived and -realized characters, imagination of concepts and rigorous exploration of them, prose style...).

        Few readers have time for more than, what, perhaps a hundred novels a year if they're eager? You could easily fill that with the winners and runners-up for various awards, say, without ever touching on the ones that didn't get a look in.

        We can regret the talented authors and terrific novels we'll never hear about, much less read. We can celebrate that we have this abundance of great literature available to us, so if we allow ourselves to be even a little ecumenical in our tastes we need never run out of wonderful books to read.1 Or both.

        Will a large increase in the number of machine-generated novels make it harder to find the good ones? Perhaps. Personally, I keep finding that recommendations from reviewers and friends and the local bookshop and even, astonishingly, Amazon point me to new (to me) authors that I enjoy tremendously.

        1And that's if you're not already an incorrigible re-reader, like Jo Walton is (or I am, for that matter).

  11. Anonymous Coward
    Anonymous Coward

    Life’s too short

    To waste time reading bad novels.

    Same goes with drinking cheap wine.

    1. Fifth Horseman

      Re: Life’s too short

      I'm not saying it is actually a bad novel, but James Kelman's "How Late It Was" makes much more sense after a couple of bottles of Buckfast.

  12. ChoHag Silver badge

    > Bots can repeatedly reach out to recalcitrant borrowers without getting fed up or burnt out by abuse.

    This might make customer service great again, just make sure you're not talking to a real human first (or if you are that it's the boss: you can claim dyslexia).

  13. philstubbington

    GIGO

    Having read just the blurb of one of these, they’re very poor quality.

  14. Charlie Stross
    FAIL

    Nope, this is nonsense.

    > Each book in his "AI Lore" series is between 2,000 to 5,000 words long - closer to an essay than a novel. They are interspersed with around 40 to 140 pictures, and take roughly six to eight hours to complete, he told Newsweek.

    Riiiiiight.

    Boucher notes that his "novels" are short because ChatGPT has difficulty assembling a longer story line. Well, no shit, Sherlock: deep learning language models are purely statistical, so while they produce output that follows the pattern of human-generated text, there's no underlying model of a fictional universe below the surface: it's not going to give you a text where statements in one paragraph line up with the next and lead to an unrolling story with developing dialog, character growth, and plot.

    He also claims to have sold 574 copies out of 97 works in a nine month period.

    For comparison: my latest book dropped last week. I'm not a bestseller, but it's made around 3000 sales so far in the UK, and probably 4-5x that in the USA. So probably around 12-15,000 sales so far, from one title. (Yes, publishing is an artisanal business with vastly smaller audience figures than you might expect if you're used to film, TV, and computer games. Nevertheless some of us do make a living at it. Just, not from Boucher's level of output or sales volume!)

    I have no definite insight into why Newsweek focussed on this guy but I suspect it's astroturf product placement by the AI hype merchants. In terms of being an actual threat to real working SF authors, this horse has been flogged to death.

    1. vtcodger Silver badge

      Re: Nope, this is nonsense.

      "I have no definite insight into why Newsweek focussed on this guy ..."

      ChatGPT told them to? (In fact ChatGPT might have written the article. How would we know?)

    2. Michael Wojcik Silver badge

      Re: Nope, this is nonsense.

      I broadly agree (per my post above about the human-cognition multi-agent model; it might be possible to make a transformer model big enough to simulate all those processes, but it would have to be much bigger).

      That said: A recent paper explored a method for making the transformer compute cost linear, rather than quadratic, in the token-stream length. If that works out, it would be conceivably possible (albeit expensive) to have a context window of, say, 50K-100K words, which would let the LLM operate over the entirety of a novel draft and maintain logical consistency. I doubt the output would be particularly interesting – there would almost certainly be the occasional surprising gradient, but the parameter space for an LLM trained on a large, diverse corpus necessarily tends toward minimizing information. But size of the context window alone appears to not be an inherent limitation of LLM architecture for something novel-sized.

    3. Martin-73 Silver badge

      Re: Nope, this is nonsense.

      We are amongst royalty, if this is the Charlie Stross I think it is!!!

  15. Anonymous Coward
    Anonymous Coward

    Ironically enough, many years ago, a real sci-fi author, Arthur C Clarke, wrote a story about exactly this happening.

    1. Ken G Silver badge

      I believe you but can't remember/find which story you mean. Can you tell us?

      I've read loads of Clarke, when younger, but he was always a better futurist than an author (great ideas, weak characterisation).

    2. Charlie Stross

      Not Clarke, but Fritz Leiber's The Silver Eggheads from 1961 (link goes to US kindle store: currently out of print on paper and in the UK) fits the bill.

      1. This post has been deleted by its author

  16. anthonyhegedus Silver badge

    chatGPT for iOS is only available in the US for some reason

  17. Jedit Silver badge
    Headmaster

    "Boucher's superhuman output"

    Excuse me? Back in the late 50s/early 60s Lionel Fanthorpe authored 89 158-page novels in a three year period, dictating them onto tape and having them transcribed by a typing pool. They weren't good novels, of course, but then I doubt what Boucher has put out is any better and even at 5000 words each Boucher's "superhuman" output would be scarcely a quarter that of Fanthorpe's actual human output.

    Other cases of extreme speed writing include Michael Moorcock writing the four-volume History of the Runestaff in twelve days and Robert Louis Stevenson writing The Strange Case of Dr Jekyll and Mr Hyde in three, destroying it at his wife's insistence, then rewriting it from scratch in the same length of time.

  18. Martin-73 Silver badge

    "indie"... lost me at that point, if an author says that kind of thing, it's probably not worth it

  19. HMcG

    El Reg, routinely criticising the LLM / AI hype train, while also jumping onboard and yelling “Choo-choo” as loudly as they can. :-(

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like