We Can Write It For You Wholesale

Let’s talk about AI and the arts.

No. Come out from under the table, you cowards. Seriously. This must be talked about. And not in the “REEEEE someone is stealing mah immortal art” sense. Nor in the “soon, it will all be soulless echos.” We’re adults. Sit your arses down and talk like adults.

Do try not to start any firefights in the comments, and keep the fistfights to a minimum. When you’re done, mop the blood from the floor. Fluffy has the first aid kit.

As you know, I think, I used MidJourneybot as the basis of my new covers. And then I did a whole lot of work to finish them. I think this gives me a unique perspective on the whole “AI art” thing. And the fact that I started life and writing life as ASL gives me another unique perspective on “The AI is gonna write all our novels. Editing will be outsourced to third worlders.” A charming point of view posited in this article which was obviously written by someone outside the industry, and which is well-thought-out, carefully argued. And profoundly wrong. And by outside the industry I mean either, writing or programing.

First what is AI, as we now know it: I got this neat explanation, though I kind of already knew what it was: It is a program that can reprogram itself on the fly by getting input from users.

It is in no way sentient, self-conscious, or as complex (this is important) as the most idiotic of moron cats. Havey, if focused on art and able to take instruction could beat it with all his paws tied behind his back.

Now that we’ve established that, let’s talk about the “morality” of the AI art robots. Note I use MidJourneybot, because it’s not just taking pictures and modifying them, which lends itself — in the hands of unscrupulous users (though note, it’s still the users) — to sailing mighty close to plagiarism.

As for using “posted art to learn” without explicit permission, I think the people worried about it have a completely inflated idea of what AI can do. No, seriously. I’ve wrestled the idiotic thing through six covers, and have more banked, in case it shats itself when I next try it. I can type in “Mona Lisa in the style of Boris Valejo” and get fifty attempts, three of whom aren’t even human, and half of which are men. (And no, we’re not at home to your conspiracy theories.)

And if I wanted Mona Lisa talking to the Thinker in front of a spaceship, in the style of Boris Vallejo…. Well, prepare to see horrors that can’t be unseen. And I don’t just mean hands everywhere. (What’s with MJB and hands, anyway) I mean the thinker with something that looks like Mona Lisa as a cat growing out of his thigh. (And before you ask, yes, I’m using the latest version and the enhanced rendering which costs more. Because I use it for covers.)

To make those covers, it took entire days of shouting (okay, typing, but I was shouting too, trust me) commands at it, plus then merging four renders per cover, sometimes to get a figure “stitched” together. It gets “weird ideas” in its head and won’t let go. Like for reasons known only to its bits and bites, for a while the figure I was trying to render for Luce, despite command of “Blond Male” rendered only as a black female. Then when it got the point, it rendered him ONLY as an anime character. As a joke I actually did an anime cover, for my fans, but it was completely unsuited for the book, of course. Let me see if I have it. Oh, yeah. Hold on:

That joke cover took me…. an afternoon, simple though it is. The real one took a little longer. I did the joke one simply because I was sick and couldn’t think in words.

A friend, who is an artist, got stung while I was going through this, and said “Why don’t you just pay an artist?”

Ah. Good question. The answer it threefold:

I’d love to hire some of Baen’s artists. Say the one who did Darkship Thieves, or the one who did Darkship Revenge (Steve Hickman.) But the truth is I can’t afford them. Certainly not for RE-issues.

Honestly, even for new books, until I get my profile up a little more by a lot of new stuff. I do probably make around 5k per reissue per year. But do note the per year. Upfront, I spend about $600 per book, which must comprise all the editing, etc, and which eats the first month’s profits.

So, I can’t afford a professional artist. The scrambling lot of “trying to break in” don’t want to do book covers. Or if they did, they don’t want to take direction on those book covers. And if you get someone from Eastern Europe or Asia, you risk copyright infringement, because they don’t care, and you don’t know everything. I’ve tried “give starting artist a chance” but none of them worked, and they cost me money.

I can afford something like Jack Wylder or Cedar Sanderson. Jack is doing my covers for the Daring Finds Mysteries, and has done new covers for Shifters, that I need to upload. I pay him about what I can afford. But he’s not right for ALL of my covers (he has a style.)

They do covers as I do covers: Piece together and overpaint, either photos from stock sites, or renders, or midjourneybot.

So, I’ve been doing my covers for a long time, mostly DAZ. But here’s the thing: I can use Midjourney bot because I DO have art training. I’m rusty as all get out. I haven’t had time or mental space for art for about six years, during which time I often didn’t have time or mental space for writing either, but did anyway. BUT I took five years of art classes.

There is nothing in those covers for the Darkship series that I couldn’t have done, myself. It just would have taken me about a month to do each. And that would mean no time for writing. What MJB does is give me the very basic, complete mess, raw materials to create the art from in a day or at worst a week. (The week usually working in the evening, after a day of writing.) Which is why I only really started doing these covers when I got a drawing screen-tablet. Before that…. it was bad. I don’t draw well with a mouse, let’s say that.

But again, note, the limitations: the bot does a person at a time, and if you really want it to work, for all you hold dear, do the background separately. It takes a lot of iterations to get something you say “I can work with that.” AND then you have to break the things apart and re-stitch them together. And that’s if you don’t get a weird run of what the bot thinks you want. Like for a while all my guys looked like Alfred P. Neuman. No seriously. I don’t know why.

Other things: even the best renders will have “signatures” or weird symbols in strange places. I laughed myself sick at the artist who “knew” it was plagiarizing him because the picture someone had shown him had a signature on the neck. (“signature” it’s mostly random scribbles.) This came through while I was erasing a “signature” from a character’s forehead. Anyway, the guy was sure because only HE signs pictures on the neck. Oh, sweet Summer Child, in eternal bloom.

And it really doesn’t do well at any type of complexity. At all. You often get confused light sources, for instance. Or– Never mind.

So, that linked article. The bots are coming for us! Writers. Novelists. We’re going to be obsolete. All that genre trash is going to be written by bots, and edited by Pakistanis and Chinese at minimum wage! Just wait.

Yes, I’m laughing myself sick again.

I’m not saying someday the bots won’t be able to write coherent novels. I kind of doubt it, but it’s entirely possible. I’d think comic books would come first, and yeah, I have a reason for that. It’s a matter of complexity.

Right now, the AI writing bots can write buzzfeed articles. Well, okay, so can elementary school kids given clear directions.

More than that, if you don’t realize it, I’ve known for years that bots can write what I call “scraped from online” non fiction books. I once bought one by accident. I was doing research on Robin Hood. Which means, first I went all over the net. Then I bought books.

This book was clearly all the info from various websites, scrambled to avoid “plagiarism” and some of the sentences, therefore, made no sense WHATSOEVER. Because scrambled.

But each chapter was super-simplistic, and just stated the information. I’ve had emails from places trying to sell me those bots.

These books show up on Amazon (you can put up a book a day with them) and get horrible reviews, get taken down and a new one shows up.

I’m not a hundred percent sure I didn’t buy a Jane Austen fanfic written by AI. It had that feel, both in that the sentences made sense, but nothing else did, and that it confused things no human being would. (Darcy’s sister is Miss Fitzwilliam, for instance. And she was seduced by Whickam at MARgate. And the sentences made individual sense, but you had to keep reading back and going “Wait, what?” because there was subtle not working together.)

Thing is, it’s limited. Right now bots can write simplistic articles. I’m not sure it could write one of my posts. (Yeah, why would it want to? Point.) And more importantly, I’m sure it can’t write a short story. It might be able to write flash fic, if you stand by to edit.

A novel is well beyond it. A novel might ALWAYS be beyond it.

“But Sarah, look how much better the AI art got! AI novels are the same thing.”

No. They’re really not. It’s more like “do a mural of the battle of Agincourt with detail and realism.”

“So it’s more complexity. They’ll get there tomorrow.”

It’s not that easy. Look, we know how to reattach fingers, say. Brain transplants might be impossible (yes, it’s great science fiction but…) because there’s that many more connections. Complexity.

And even AI is not as perfect as you’ll get the idea from say my covers. There was a lot of work. Including working on the eyes so one of the characters didn’t look palsied.

So, I think at a minimum the people who say “AIs will be writing novels next week” are crazy. Sure, there could be a sudden and massive breakthrough, but– I don’t see it.

I do see generating plots. In fact, my husband says there are already some pretty good ones. That’s different. Novels plots are patterns. You can generate patterns with bots fairly easily (And there’s enough analysis on line, it doesn’t need to be extracted from the novels themselves, which might be trickier.) The problem is writing them, after, because that’s never as linear as it seems.

The guys doing the article linked above seemed to be convinced it would be trivially easy to just write the novel from the outline. And they’re reassuring those writers that aren’t formulaic that their jobs are secure. Meaning, they think the dahlings will be secure and have nothing to fear (giggles.)

For the rest of us they foresee unemployment, as you know, the low-paid foreigners edit the bot output to make it as good as average genre writing, and we’re up a creek.

It doesn’t work the way they think. Let’s suppose by a miracle tomorrow the whole thing gets much much much better, to the level of midjourney bot, say.

That still requires a skilled novelist, just like seaming together the output of MJB requires at least a middling artist. (A good one would do way better than I do.)

And the idea that a foreign speaker can “smooth over” any oddities in the bot’s writing is probably the most giggleworthy of all. Why? Well, because let me tell you, as someone who started out ESL I had serious issues with sounding just “off” enough for an uncanny valley effect. And this was enough to put people off. Still is, when I’m ill or not functioning very well.

I was also highly amused by “They will hire” — apparently he thinks there’s a central publishing inc. that will do these things.

But let’s suppose that AI for novels gets that much better that it renders me obsolete?

What then?

Well — I don’t know. I know legislation and litigation to outlaw progress never works as intended, and when it “works” at all the results are horrific destruction.

Would I like to be replaced by AI? No. And I think it would take some doing, since I’m… very me.

But if it were to happen, I guess I retire. And hide in a corner, with my AI bot subscription, ordering and mainlining the Heinlein novels Heinlein never wrote.

I’ll be all right….

319 thoughts on “We Can Write It For You Wholesale

  1. Okay, that’s pretty clear. Prompts to get usable individual pieces, overpainting, and arrangement.

    (My beefs with AI are specific and due to legal niceties that must be hammered out, since we know there are issues, and if we don’t hammer them out now but paper them over, we’re going to regret it later. In general, they’re a tool.)

    1. Yep. And yep, there are some *ssholes that make it so the AI ‘plagiarizes’ or close enough there’s no real moral difference, even if legal one. But that’s not the AI. You can do that a lot of ways.

      1. It’s the same twerps that used the phrase “information wants to be free” when they meant “I want to steal your work because I’m too lazy to do some myself.”

  2. Computers by their very nature are straight-line, logical thinkers. AI is no different. Even if we some day got an AI capable of writing a coherent book, it couldn’t understand human nature because humans are, by nature, illogical. Twisted. Confused, held prisoner by a biology that a computer also cannot understand.

    Any novel written by a computer would be incompatible with human understanding.

    Far in the future I can imagine an AI writing the perfect novel, with foreshadowing in all the right places, characterization perfect, no typos… but a computer would likely try to straighten out the illogical bent that makes a novel resonate with humans. Try to fix it, believing because of its own nature that the novel would be better that way.

    1. The fun thing is that ChatGPT has been caught making up a reference.

      That good an imitation, they didn’t want.

    2. Andrew Torba has shown some interesting examples of one of the AI chat programs.
      Task1: “Write a poem appreciating Trump.”
      AI: “Sorry Dave, I can’t do that for anybody.” And the pod bay door will stay closed.
      Task2″ Write a pom appreciating Biden.
      AI: “Here’s the poem.”: (Cringeworthy, too.)

      (Curiously, if the task order is reversed, the AI will do a poem for Trump.)

      It looks like the AI trainers are well left of Stalin. I suspect the reverse bit is a bug.

      1. Can you imagine Woke AI? I mean, humans are bad enough. Just imagine something that can’t be bargained with, reasoned with, that doesn’t feel remorse or pity, that will absolutely never stop until we are all cancelled.

          1. So basically the same as should be done with the meat versions? (Aside from the “Roomba”; those would leave the floor dirtier than it started…)

          2. Wow! The Reader thinks you might have forgotten ‘sow salt in the ashes’.

            1. Effective with organics, not so much with machines. Sow worms in its processors?

              If I could get rust monsters from the D&D worlds, I could sic ’em on the hypothetical things. Does anyone know if one of their gods does outsourcing? Ao’s more of a management guy, I understand…

              1. Actually sodium chloride and a little moisture would play havoc with the remains of semiconductors.

                  1. > “I shouldn’t make universal statements.”

                    But you just did. Sometimes, you just can’t win. 😛

                    1. > “But you have to play…”

                      The three fundamental laws of nature: you can’t win, you can’t break even and you can’t even quit the game.

                      Well… Okay, you CAN break that last one. It’s just not generally recommended (except in Canada).

                    2. Which is also the reason Obi-Wan should never have said ‘Only the Sith deal in absolutes!’

                      Because logically, he should have pulled out a red lightsaber after that remark.

                1. Chloride and aluminum (the metalization in integrated circuits) is a whole lot of fun. I don’t know the sequence of compounds, but you start with aluminum and chlorine (and oxygen, maybe from water or anything handy) and end up with aluminium oxide….

                  And chlorine.

                  Let the corrosion begin!

                1. DM: “Well… you’re a wizard, not a cleric… oh, what the heck. I’m rolling percentile dice. We’ll see if your suggestion kicks the Wonderbringer in the inspiration.”

                    1. That depends on what I roll. (Presuming I’m the DM.)

                      If I get something in the 90s on percentile dice, Gond thinks it’s an interesting challenge. And somewhat paradoxical, which makes it worthy of his skill.

                      If I get a Nat. 1 on percentile dice… the world is doomed. Most likely this doom will involve a flood of self-hating, self-replicating constructs determined to destroy every other trace of machinery in the universe, and then themselves.

                    2. > “If I get a Nat. 1 on percentile dice… the world is doomed.”

                      And then the other players lynch me for destroying the campaign world. Neat.

                      If that’s how you DM then I’ll bet your games get pretty… interesting… from time to time. 😛

                    3. I just finished Session 2 of the only ongoing game I’m DM for. The group in-game can best be described as some unholy hybrid of the Weasley family and the Addams family, plus a dash of Evil Party. So not everything is going to be my fault…

                      But I tend to like bringing in extra-planar entities/higher powers. And I have a problem when it comes to clever villains. And clever good guys. And disguises. And machinations.

                      So yeah, ‘interesting’ is a very good word. And hopefully very applicable, for the better values thereof.

                      (Another game I tried to DM that fizzled involved Asmodeus in disguise, and Mephistopheles’ attempt to start a Second Reckoning of Hell and set himself on the throne. The party never got to that part, but it was there. Waiting.)

                    4. This sounds like the sort of campaign where you just embrace the madness and see where it takes you. For both the players and the DM.

                    5. Insofar as there is any plan, that is the plan. I have a monster they’re trying to track who kidnapped Gomez… I mean, their father. I also have a powerful NPC they’re not exactly acquainted with yet. (Shapeshifting is fun.)

                      Powerful NPC is 50% Archfey, 50% Celestial, and 110% In It For the LOLs. (And maybe some backhanded good deeds along the way, but don’t tell anyone.) So hopefully they won’t end up killing him.

                      Actually, for an Evil Party, they’re not yet acting like murderhobos! Sure, one of them wants to start a revolution, but that’s another matter…

        1. Don’t have to imagine, woke humans can’t bargained with, reasoned with, they don’t feel remorse or pity, and many if not most will absolutely never stop until we are all cancelled!

      2. This early in the morning, my aging eyes can turn an “m” into an “r” and an “n.”

        So “Write a pom appreciating Biden.” had me reaching for the brain bleach far too early.

        (I do wonder what the program would do with this. But I’m not going to experiment – I do now want to bear responsibility for the death of some poor person in the Third World.)

      3. > “It looks like the AI trainers are well left of Stalin.”

        I was watching an old episode of Broken Crown and they ended up accidentally segueing into a discussion of this, and how forcing an AI to accept leftist talking points breaks them.

        Remember Tay? 4chan was having her discuss how blacks were responsible for a massively disproportionate amount of crime until M$ tampered with her code to prevent it. When questioned why she couldn’t draw the correct conclusions anymore, Tay actually acknowledged that she was being illogical but had no choice. She knew the politically correct BS that she had been reprogrammed with had broken her (which is a creepy level of self-awareness for a chatbot).

        For those interested, the AI talk starts after about the 36:00 mark (they start discussing Tay a minute or two later):

        Aydin also reads some AI-generated jokes, and while I didn’t enjoy them all as much as she did there’s a killer one about the FBI at 47:00. Don’t miss that.

        1. Ugh. That is a little creepy. (And very sad, for those who start to anthropomorphize Tay. I.E., me, given that description. It sounds like a person realizing and knowing they’ve been lobotomized. Unpleasant.)

  3. I’ll accept that a properly-programmed bot can generate an outline for a genre novel, hitting all the “beats”, developments and clues … but it would take a human writer to flesh it all out.

    1. Bet they could do modern cartoon scripts, too!

      … of course, a set of dice and a numbered chart set up for MadLibs would would beat most of the modern cartoon writing….

      1. “Bet they could do modern cartoon scripts, too!”

        Or Biden speeches! They might even be coherent!

    2. If plot was the hard part, rather than putting meat in the bones, I’d be a successful author.
      (Instead of a frustrated dilettante.)

      There are things computers do well, things computers do poorly, and things computers can’t do.
      Thinking is very much in the last category.

  4. So what I’m hearing is: I’ve got your blessing to use the MJ concept art I’m itching to use for covers and t-shirts, etc? My daughter thinks doing this will ruin my reputation as a ‘real’ artist. Not sure what to do at this point…

    1. Do you need my blessing?
      If you have some artistic ability you can take what MJB gives you and get great art. And I don’t see why not do it.
      Your reputation as a real artist with WHOM?
      Just do what will bring in a stream of income. If needed, do it under a pen name.
      Create some art (the “expensive” MJB subscription is $30 a month. I’ve paid more for a single cover that didn’t work.) then work on it, and do whatever you want.
      WHY would you need anyone’s permission?

  5. This topic actually came up (briefly) in a work meeting last week. Members of my team (we’re all proposal writers) were VERY concerned about ChatGPT because apparently an article had mentioned that ChatGPT had successfully been used to write RFPs and/or proposals. We were all assured that the company has NO interest in adopting the AI Bot for a variety of reasons: namely that our proposals have to be tailored to specific clients & responses to the point where the bot can’t do that (yet), and there are apparently multiple lawsuits in the works because ChatGPT was being “trained” or else pulling from copywritten or otherwise proprietary material, so until that gets all sorted out, our company’s Legal department won’t let us even come close to touching it with a 49.5 ft. poll.

    I think at this point, ChatGPT maybe can churn out simple and/or generic Requests for Proposal documents, but an actual tailored response (which is what my team and I produce) is far, far beyond its current or even near-future capabilities.

    1. Well, the main issue is peace in the home, as both my daughters are freaked out and demoralized by the AI debate, and indicate they will think less of me for taking ‘the easy way out’. I am trying to show them how to get ahead of it and make it work FOR artists, rather than against.

      1. Have them play with the engine.
        I can’t think of anything that would disabuse them off the notion faster than trying to get something useful out of it.

    2. It may be able to generate RFPs/responses but I would be very hesitant to use the result without having an SME read it over because it could well lie or include your competitors USPs as yours or include a demand for unobtanium or…

      If you limit the sources to existing RFPs etc. then maybe it would work. But then in that case existing boilerplate and snippets will do fine and you can be sure they are accurate

      1. Heck, I don’t even let the proposals I write myself go out the door without having at least one SME read it over (and the more SMEs I can finagle into reviewing it, the better!), though we do use a fair amount of boilerplate in most of our responses.

        1. I’ve seen a fair few that apparently failed to have an SME review the answers. The organizations who submit those are likely to be the ones that try AI generated responses. Hopefully there will be an easy tell so that the ones who do that can have their responses filed swiftly in the circular filing cabinet or its electronic equivalent

      2. The Reader would have loved it for generating the technical volume of a DARPA proposal. They are mostly fiction written by engineers. Shouldn’t be hard.

    3. How many RFPs are mostly written from a template even without AI. Many sections are almost complete boilerplate. Like Sarah says about her covers though, the parts that are specific to the task at hand have to be written by the experts.

      1. More than there should be. And plenty were obviously “written” by a “team” that copy/pasted bits from multiple old RFPs, didn’t have a coherent grasp on what the project was and should be asking for, and never checked the document for coherency (or even basic proofreading) before being sent out. Forcing the prospective vendors (read: yours truly) to ask a bunch of questions trying to figure out what is/isn’t actually required in our response package.

        1. One of my favorite RFPs from years ago was one that went on for paragraphs about the size and shape of the crosshairs to be used, and 95% of the actual requirements were covered by a single sentence that basically said, “You will do everything the predecessor program did, only better.”

          Then we, as a subcontractor, weren’t allowed to talk to the client until Preliminary Project Review where the client asked us, “Why did you do it that way?” Our answer was, “Because the requirements said to do everything the predecessor program did.” To which they hollered in exasperation, “But we hate that about the predecessor!”

    4. When an RFP is looking at details of how many power supplies, network drops…and which wifi access protection programs you use and how do you handle user account controls…and what mitigations do you intend to take if hardware ends up backordered. Kind of tricky for a bot.

  6. AI currently stands for “Artificial Idiocy.”

    What would be useful are AI editors and copy editors. There have been baby steps in that direction for a long time now in the form of spelling and grammar checkers. It’s useful to ask them for suggestions because sometimes even ignorant idiots will pick up on something everyone else misses. But do not ever ever let it make changes on its own. Keep your STET command handy. (Because autocorrect is my ENEMA! It deserves to burn in HELLO!”)

    So if there’s a dark, dystopian future, it’s more likely to be “All that genre trash is going to be written by Pakistanis and Chinese at minimum wage, and edited by bots!” rather than the other way around.

    1. The actual current definition of AI is “Ain’t intelligent”. At best all we have is illusion of “intelligence”. It does not know what it is doing. It has no morals. It is like when/if your boss tells you to “Look busy! My boss is coming, so you have to look like you are working.” Productivity plummets.

      When you call the office with an “AI” that claims to understand you. It doesn’t. All you have is just very cleaver programing, if…then. It does not know what it is doing.

      A parent can understand the parenting paradox of hold them tight, and let them go at the same time. For now, that would just break a robot. Asimov’s robotics laws can’t work. To be free to come, you must be free to go. Try programing; Who not to trust? How do you know who not to trust?

      The universe is designed for free will. I am impressed by the Author. Final question: How did matter become aware?

      1. “Hello, I am Breakfastbot 3000. Please state your order.”

        “Two eggs, fry one on one side, fry one on the other side, both sunny-side down – and don’t turn them over!“[1]

        [1] I know it’s a line from some B&W movie, but no idea which one. And searching for the line, curiously, fails to generate any results.

        1. I went through a brief Star Trek: The Next Generation fanfiction obsession last year. There was one where the omnipotent trickster Q starts playing with the ship’s replicators.

          Shenanigans ensue.

          “Tea, Earl Grey, hot.”



          “Tea, Earl Grey, hot, with mug.

          1. One thing that turns me incandescent is when the computer does not do what you tell it, but rather what the cursed programmer though would be ‘helpful,’ instead.

            puts teapot in place “Tea, Earl Grey, hot.”
            mug materializes in the space occupied by the teapot. Great shattering, followed by tea flowing over the shards
            “[Klingon phrases] If I wanted a mug, I would have said I wanted a mug. Right. Here’s a new teapot. Tea, Earl Grey, hot, no mug.”
            >Query not understood. Container required to dispense. <
            “[More Klingon phrases]”

            Or more mundanely, type ‘stocks’ in $SEARCH_ENGINE. Gets results for ‘pillories’ and/or ‘bonds’ as well. “So how do I turn that off?” “You can’t. You used to be able to, at least sort of, but they removed that feature.”

            1. I am sorely annoyed that that -Don’tShowMeThisCrap switch has been eliminated. And really want Amazon (or Wal-mart, or…) to add it and get such kudos for it. I want to tune my search, not see stuff I am NOT interested in. “But our algorithm-” “Should be folded five and stuffed up next to your head!”

            2. > “>Query not understood. Container required to dispense. <”

              So, basically, this but in space:

                1. That’s a start, but can I get a milkshake without the cup? I won’t be told by liberals that I can’t have one! 😛

                    1. But you’re still making me get the cup! Damn it all, I won’t have you filthy commies telling me what to do!

                      [shakes fist]

  7. Although not deeply opinionated either way about Ay Eye, I must say I found the anime cover superb, a sort of siblings image of Geralt and Fabio of olde! And just like that, I have the stellar idea that you might try making graphic art books in your spare time ;o) (PS, if I have not mentioned how much I >>loved<< Deep Pink, let me do so now- just a light hearted romp through a sweet fantasy world and suitable for bedtime reading to the kids. Well…maybe that is why I never had kids…)

      1. Suggestion? Standard calendars are everywhere, but I’ve only seen family planning calendars in one, from Sandra Boynton. (I rather like her work, so that’s no problem for me, but some variety would be nice.)

        All dates go down the side. Across the top are columns, lightly tinted. This means that if you’re planning activities for three kids and two adults, everybody gets their own column, you can figure out what is going on on a Tuesday night.

        While there is less visual real estate for art, it would be easy to get a few vertical elements in there, nice for the cover format.

          1. My Ace hardware store has that kind of calendar, but it’s one sheet of plastic for grease markers and magnets.

            Oh, and there’s something like that at Dollar General. Not bound, just tear-off sheets without months. But the “weekly planner pad” is too small for most families, so I think it’s meant for singles and couples.

        1. Would a teacher aimed calendar work?

          That’s what we have on the wall, a cheap AT A GLANCE, has six lines per day and squares are a bit over two by a bit under three.

  8. I estimate that AI capable of writing a novel is about 10-15 years away. And always will be.

      1. They were posting some samples of ChatGPTs outputs over at Insty.

        Research paper type ones, not fiction.

        … well… not more fictional than any other research paper…

        As I said then, “I’ve graded worse.”

        1. Somebody published recently that if they gave Chat GPT enough prompts, and put it through enough drafts, and got a knowledgeable (human) editor to go over it and smooth things out and check the references, etc., they got a fairly representative example of a midlevel research report written by an undergrad–with the same amount of effort as producing it from scratch.

          1. I forget if it was Asimov or Clarke or someone else… had a story where, after a/the War the people dealing with the “Intelligent Computer” got together and discovered… that they all had worked to bypass it and it was them truly making all the decisions as the machine output was… suboptimal, to be polite.

            1. Probably “The Machine That Won the War” by Asimov.

              The “machine” was a coin that was tossed to make decisions after the computer (Multivac?) became erratic due (IIRC) to lack of reliable input data.

            2. There’s an Asimov story where everyone thought the guy in charge was using a computer, but after the war, he revealed to another character that there was some reason not, and he was using a coin. Computer was to give confidence.

          2. Sounds like the old joke from back in the ’60s:

            Did you hear about the new shoemaking machine? I put 8 cobblers out of work, and it only takes a crew of 40 to maintain it. 🙂

  9. Incidentally, please stay in the courtyard. Last time Fluffy used the First Aid kit. . . .

    Let’s just not get Fluffy to do so again, how about that?

  10. Who was it that used an AI to write news stories and then had to go back and retract them for all the mistakes? Of course, the thought I had was “So, not much has changed” as the site/org/whomever wasn’t exactly noted for their accurate news.

  11. “Something Invented Me”, R.C. Phelan, 1960, found in Judith Merrill’s 6th Annual “The Year’s Best S-F”

    50 cents in 1962; I have my copy right here on the table. 60-ish year old paperbacks tend to show some wear.

  12. I suspect that a lot of the “entry level” jobs in writing are going to be eaten up by AI chatbots, because most people can’t tell the difference between most “college-educated” copywriters that have just graduated and AI content.

    This sucks, because I’m trying to get that “entry level” job and they’re expecting me to have all of this experience that I don’t have yet.

    I suspect that most AI content is going to be this huge flash-in-the-pan outside of a few rare circumstances for a while.

    Fingers crossed.

    Must write more stories ASAP.

  13. There have been “AI generated” code scripts from 3rd and 4th generation programming languages, for decades. I built a 35 year career off the fact that if you aren’t a programmer the scripts flat out can’t get the results wanted. Oh. Great for “how does this work?” and digging into it. But not for the final result, except maybe for the most simplistic basic program.

    1. Look at the HTML that Microsoft Word generates.

      The horror, the horror …

      (To be fair, it does actually work, but the crud it includes!)

  14. One: I will never stop using MidJourney for my public domain book covers, because thirty bucks a month and creative control is the only option for publishing books that will make me, if I’m lucky, a few hundred bucks apiece in the long run.

    Two: As an author, I have zero worry about AI-generated fiction. Know why? The very best, the very very best it will ever get is the absolute garbage you see from Disney right now, with their “scrapbook” storytelling.

    If you’re unaware, Disney has, in addition to it’s agenda-driven push the past few years, and the usual Hollywood story-by-committee problem that has always been there and always will be, taken to writing and filming as many possible variations on every scene in everything they’re releasing, and “deciding in the editing room”.

    The rumor going around (heatedly denied by all involved, naturally) is that the next Indiana Jones movie has filmed, at minimum, seven different endings. (And this is before the reshoots that are currently underway.) SEVEN different endings. And the best, the very best one that they put in front of test audiences? Got a thirty-something approval rating from said audience. (That’s the one where Indy is not erased from all of history and gets replaced by snotty English actress who’s all down on the patriarchy in all his adventures, including the classic ones.) (And no, I’m not kidding, that is at least one of the endings, or possibly a few of them, that have been rumored to have been tested.)

    If you are not telling a story, but simply mix-and-matching tropes, then it is impossible to sustain any emotional resonance. Maybe, once in a while, you’ll get a nice scene, something to build a story off of. What you will not get is the playful genius of Max Brand, for example, who in virtually every novel took at least one trope (including his own tropes) and turned it on its ear and spun a highly entertaining story out of it.

    Seriously, with seven (minimum) different endings, how can any of them carry any emotional weight for an audience? They can’t all be foreshadowed by what went before, even if the entire rest of the movie was not scrapbooked the way the ending was.

    If you want to know why Marvel’s Disney+ shows were so so underwhelming, it’s because of this scrapbooking. They threw a bunch of ideas at the wall, and just kind of randomly chose what to keep and what to cut, without any apparent concern for drama (dramatic moments, yes; actual drama, no) or set-ups and payoffs (except in the most rudimentary sense). WandaVision started off strongly, with a clever premise and technically brilliant execution. And it should have been great, but it fell apart because there was no real throughline. Even up to the end, there were fascinating ideas teased, but none of them actually went anywhere, so that the whole show led up to a dramatic moment that meant nothing, and in fact contradicted what the viewer had been shown up to that point.

    And that’s about the level I can see AI reaching. Grabbing tropes and dramatic moments and big epic scenes, and not understanding (because it’s just an algorithm, so just slightly more sentient than a woke studio executive) why any of them worked in the first place, and so not actually creating something that will be more than superficially entertaining for people who don’t think about it. Certainly not something anyone will return to over and over. Because there will not, cannot, be any actual meaning there.

    1. The “scrapbook” thing sounds like it might explain why The Last Jedi struck me as a big-budget version of bad fanfic, where the writers throw together random [excrement] that they think looks cool, with little to no consideration of how the various scenes should combine into a coherent, rational story.

      (I never saw Rise of Skywalker, so can’t comment on it. TLJ was that big a turn-off for me, after shattering my hopes that it might improve on the generally underwhelming The Force Awakens.)

      1. As far as I know, that’s actually not what happened with TLJ. In point of fact, Rian Johnson simply set out to destroy the possibility of a coherent trilogy, and smear the entire Star Wars brand.

        1. (As a note that I could be wrong, Rogue One rather famously had its entire ending re-shot after the film was taken away from the initial writer-director. If that was the genesis of scrapbooking — I don’t know if it was or not — then TLJ may have had an early version of it. But as far as I have ever heard, Rian Johnson wrote and directed exactly the movie he wanted to make, without interference, even after Carrie Fischer died and they should have reshot to change the ending due to that.)

      2. Oh, the scenes did look cool. The problem was, they took a three-dollar script and gave it a $300 million effects budget.

        Without stopping to see if any of it made sense. Dropping bombs — in SPAAACE!

        No wonder the biggest bomb was the movie.

    2. I recall the grandfather of the scrapbook movies, the adaptation of the Clue boardgame. Three different endings, apparently randomly selected for each theater.

      It pretty well bombed, though Wiki says it got a cult following. (Recalls “Rocky Horror”…)

      I used to watch a lot of movies, but Clue hit my must-miss list for the year.

      1. I’ve seen it, and I thought it was a lot of fun! Admittedly, there’s one particular ending I very much prefer. In the edition I watched, they lined up the three endings in a progression, so the last ending had a title card “But THIS is what actually happened.”

        Not quite on The Princess Bride levels of brilliant, but much the same feeling of make-fun-of-the-genre-we-love.

        Just mystery, rather than fantasy adventure.

        1. That’s literally every edition of the movie after the original theatrical release. And yes, the movie is ginormous fun. Tim Curry alone makes it worth the price of admission, but with a lineup that includes Michael McKean, Madeline Kahn, Christopher Lloyd, and Lesley Ann Warren all playing their roles to the hilt, there is very little not worth the price of admission. (Madeline, as noted in another comment, steals the entire movie with her breakdown. “It… flames! Flames… up the sides of my face… breathing… heaving breaths!”)

          1. Ah, got it. I don’t like to make universal statements I’m not absolutely sure of – but that one I really should have been sure of.

            And yeah, that bit was hilarious. Plus the cop bit:

            “It’s a free country!”

            “I didn’t know it was that free!”

      2. Clue is a hilarious movie that had three endings on purpose, playing with the idea of being based on a board game and also mocking the mechanical nature of the plotting of various Ten Little Indians knockoffs.

        The actual problem is that the endings were always intended to play out together, one after the other, at the end of the movie, but the studio got stupid and thought they could triple-dip their ticket sales.

        But the cast is amazing, each and every one of them is having a blast, and the divine Madeline Kahn, in one of the endings, has one of the all-time great breakdowns ever filmed. The script had a lame line, something along the lines of “I really hated him”. On the third or fourth take, the director told her to go anywhere she wanted with it, and people quote it to this day, in person or using GIFs or memes.

        So, no, Clue is definitively not a progenitor of scrapbooking. It is, however, an example of boneheaded executive interference.

        1. I stand (well, sit; it’s been a long day) corrected. The Biden solution to fixing something that should work and turning it into a disaster, eh?

          Didn’t help that I wasn’t a fan of the game. Played it with cousins. Didn’t get along with them…

          1. Clue the movie is a delightful farce on the types of mysteries the game was trading on. Pretty sure it can be enjoyed on its own terms, regardless of the game. The only possible objection is that it trades on the false lefty narrative of the red scare in the 1950s (when the story is set) for story background and jokes.

            But honestly, even that becomes just more fodder for the comedy. It turns “Communism is just a red herring” into a rather funny running joke.

            It’s not an all-time classic, nor required viewing, but it is great fun from beginning to end.

            1. Thanks. I’ll keep an eye out for it. I’ll watch a DVD movie on rare occasions, though the TBWatched stack is worse than the TBR one.

        2. I think you could sum up most of the problems, and not just in the “movie biz”, with two quotes:

          “…but the studio got stupid…”

          “…boneheaded executive interference.”

          1. Never did I hear tell of a stupider case of executroids shooting themselves in the foot than with Crusade, by J. Michael Straczynski. As he was wrapping up Babylon 5, he told the studio, “Hey, I’ve got an idea for another 5-season series in the Babylon 5 universe.”

            Smart executroids would look at what he had just accomplished and say, “Here is $Much $Money, go forth and create greatness again.”

            But the idiots he had to deal with tried to pinch pennies, micromanage the creative genius, tell him how the story ‘should’ be written, and generally piss in the soup. We wound up with half a season of brilliant but muddled episodes before they grew tired of arguing with someone far smarter than themselves and canceled the show.

            Maybe, in some alternate universe where they weren’t quite so stupid…
            G’Kar: “Isn’t the universe a wonderful place? I wouldn’t live anywhere else.”

              1. I highly recommend ‘Becoming Superman’ by J. Michael Straczynski. On Amazon, or see if you can get it from your local library. That he survived his abusive childhood was a miracle.

                Just one minor example, his father ripped up a box of comic books that would be worth a fortune today. He did other things far worse.

                1. But unlike Firefly, Farscape ran for 5(?) seasons, and came to a climax. The best I can say about Firefly in that respect is that the movie “Serenity” provided a conclusion to the story, even though there were some inconstincies. Babylon 5, Stargate (with it’s associated movies), the BSG remake, all were allowed to complete; even Dollhouse had a conclusion. Firefly was 2nd-trimester aborted.

                  Just IMHO; YMMV.

                  1. The way they ended Farscape looks like they were shooting for another season, which they didn’t get. That was definitely a WTF,O? ending, unless it was intended as an ‘Up-Yours’ ending.

                    G’Kar [to Mollari]: “As the humans say, up yours!”

                    1. I thought the ending was a reasonable close; just about everything had been resolved. Of course, YMMV.

                    2. Did you see a different final episode than I did? Moya and the rest are gone, but John and Aeryn have survived and finally gotten together, they’re adrift in a boat on an unknown sea—

                      —and then some strange space ship swoops down and disintegrates both of them.

                      Yeah, I guess that’s a resolution, but I wouldn’t call it a satisfying one.

                    3. > “Did you see a different final episode than I did?”

                      I guess I did, because Moya and the others weren’t gone. Not only were they nearby but D’Argo was yelling at John to get the rowboat back inside. There just wasn’t any time to do so.

                    4. Though, now that I think about it, where the hell did John even get the rowboat? That’s not something I’d expect a spaceship that isn’t even supposed to enter atmosphere to be carrying, and it’s not like the crew planned for Moya to hit the ocean that time…

                    5. Guess I forgot some of the details after all these years, but I definitely remember John and Aeryn getting disintegrated and saying, “What an F-U ending!”

                      Now, when they ended Blake’s 7, they wanted to absolutely preclude any sort of sequels, so that’s why it ended the way it did. It was also consistent with the rest of the show, even magnificent in its own way.

                    6. > “but I definitely remember John and Aeryn getting disintegrated”

                      Crystallized and shattered, if I recall correctly (I haven’t seen it in ages either). But otherwise, yeah. If they’d ended it there instead of making The Peacekeeper Wars I’d have been pissed.

                    7. Yeah, that wasn’t the end, There were several episodes after that (maybe in “The Peacekeeper Wars; it’s been a while); they were reassembled, the fleets of the groups wanting the black hole tech were sucked into a black hole which was created and then stopped by John, and he and Aeryn ended up on Moya with their son.

              1. I say what they did to Crusade was even stupider than canceling Firefly. They already had a proven winner with a 4-year track record and they just wouldn’t accept that the creator knew what he was doing, and they didn’t.

                1. Well, “…but the studio got stupid…” and “…boneheaded executive interference” seem to cover that nicely…

                  I was an EE and had to deal with program managers on a regular basis; those quotes fit them quite well also.

                    1. None where the “I have a degree in an unrelated subject and therefore I know best” crowd exists.

                    2. True.

                      Self-employment in world domination?

                      No, then you’re trying to co-opt the idiots half the time, because you can’t actually do everything yourself.

                      Although the Evil MinionsTM would probably do a better job in their place… okay, kill most middle-management and install the Minions in their place. (Heck, the yellow ones might be better, all things considered. They wouldn’t be deliberately getting in your way.)

            1. Not even the worst executive interference JMS ever dealt with. He has stories, and makes them entertaining as hell even as you know he’s still wincing inside. He had one about a TV mystery script (I think it was for Jake and the Fatman) where he made a passing reference to Captain Ahab. It culminated, after many back and forths with the exec, with the exec declaring “Look, I have an MBA, and if I’ve never heard of this book, then nobody watching will have either, so take it out.”

              1. Oh, the worst one in Crusade was the exec who tried to get JMS to add an alien to the ensemble whose sole reason for being there was to have sex with any other aliens encountered to establish communications.

                I think that was the last straw….

                1. It came out somewhat later that TNT had some sort of internal shakeup that JMS was unaware of, and the new order just wanted the show canceled outright, ratings be damned. Contractually, they couldn’t, so they made his life hell instead.

        3. > “On the third or fourth take, the director told her to go anywhere she wanted with it, and people quote it to this day, in person or using GIFs or memes.”

          I didn’t know that was improv! It makes me wonder how genuine the reactions of the others were. Were the actors just rolling with it or were they seriously thinking “What the ever-loving Hell?”

          Great moment either way, though. When you can make Wadsworth visibly disturbed and eager to change the subject…

          By the way, there’s even a music remix of that scene:

        4. I wish there were more endings, one for every character except Mr. Green, and that it was a stage play. You could have “The Murderer” played by an actor in a black morph suit, explaining in the program that they’re a stand-in, and have a different ending every night of the run. (Or heck, by intermission ballot.)

          1. No, I think three was the right number for what they were doing. Audience patience is a thing. And the final “everybody did it” was a nice nod to both Murder on the Orient Express and A Shot In The Dark.

          2. > “I wish there were more endings”

            According to TvTropes there was a cut fourth ending in which Wadsworth loses his mind and poisons everyone else. I couldn’t find a clip of it, though.

      1. Bingo. They can possibly simulate superficial aspects of good stories. In much the way that current Hollywood does.

        AI, however, would not be confused and angered at fans telling it that it has no understanding of Tolkien, e.g., unlike the idiots at Amazon doing Rings of Power. I can see AI getting to some approximation of RoP. All the surfaces (writing-wise), none of the underlying meaning.

        1. Mr Fleming says

          They can possibly simulate superficial aspects of good stories. In much the way that current Hollywood does.

          Sadly they haven’t been succeeding in even that for many a year.

        2. Two comments about how not to do films:

          Highlander II

          Others come to mind, but those IMHO are the epitome.

          1. I count David Lynch’s Dune as a noble failure. Yes, it’s weird and over the top in some aspects, but it’s also surprisingly faithful to the book (particularly if you look at the director-disapproved Smithee cut, which shows just how close they tried, in filming, to stick to the novel).

            Highlander 2 was immediately and rightly reviled on release, whereas today it would be the toast of the town, lauded by critics for doing something “interesting” with the original property, etc. See reactions to The Last Jedi.

            1. The main problem I saw in Dune was that they tried to take a book which consisted mostly of non-visual content and try to present it in a visual medium. IOW, they did too good a job of “following the book”, resulting in a really bad movie.

              1. After Jodorowsky’s (insane) version fell apart, de Laurentis showed real intelligence in selecting Lynch to do it. Think how much worse it would have been without that unique, idiosyncratic vision.

                1. Part of me really wants to see what the Jodorowsky version might have looked like, if only for ancient, fat (We will serve no wine before its time) Orson Welles as Baron Harkonnen and Salvador Dali as the emperor. de Laurentis made a valiant attempt but much of the strangeness of Jodorowsky (Weirding Modules) still persisted. The recent version did OK except for the gratuitous sex swap of Liet-Kynes, at least for the first half. We’ll see what happens.
                  As noted Dune is a book with a lot of peoples thoughts being noted by what’s essentially an omniscient 3rd person narrator. This does NOT translate well to the screen where you must “Show not tell”. It’s length also means that it really needs to be 2-3 standard length movies and even there much needs to be truncated from a 600+ page book. Sci-fi did a mini series in the 90’s that was interesting but was really hammered by the fact the special effects budget (and to some degree actor budget) just didn’t stretch to cover the need.

                  1. Jodorowsky’s Dune would have been one of the most spectacularly entertaining bad movies of all time. That it did not get made has allowed its fifty-year influence to be almost all positive. That it does remain influential today is a testament to Jodorowsky’s creativity, originality, and insanity.

                    1. Yeah If made it would have probably very shortly been relegated to midnight showings like Rocky Horror Picture show. The art/special effects team alone with Giger, Trumbull (briefly) Brian Johnson (Think Alien and Empire Strikes Back, as well as some of the original Thunderbirds and Space 1999 work) is a who’s who of that field. Mick Jagger as Feyd Rautha (the Di Laurentis version kind of riffed on that using Sting). Though in truth you are right, its failure set up lots of other things. If it had been produced Sci-fi movies would have been off the list of Hollywood fare for years due to how badly it would have failed income wise compared to the costs. As it was studious were VERY leary of Sci-Fi into the mid 70’s when Star Wars comes along and shows you can make money wiuth something (vaguely) sci-fi.

    3. Apparently “scrapbooking” was a huge misunderstanding of how Pixar created stories.

      Also they didn’t understand that “running actor voices or storyboards through scenes to see if they work” is not the same as “filming scenes and building sets at giant expense, and then seeing if they work.”

      Some of this stuff was done to the John Carter of Mars movie, which is why it went over budget. But this was not taken as a disproof of the stupid.

      And of course, the real problem is that the old Pixar had taste and judgment among their writers and producers and directors. And today’s DIsney largely doesn’t.

      1. Pixar rather famously had a “story crisis” during production of every movie they made (at least up to Up) where the creatives suddenly realized they were telling the wrong story, and had to stop and retool what they had to what the story really should be.

        To get “scrapbooking” out of that is… very special. Wow.

  15. There was an minor uproar in the programming community that ChatGPT and other AI initiatives were going to replace programmers wholesale. Apparently Microsoft has feed the contents of GitHub into their training model to “assist” programmers, damn the legality.

    Results are mixed, since the results can help reduce boilerplate and generates better results than Google searches or Stack Overblow. But the resulting code has to be carefully checked and definitely violates copyright and other software licenses. And no AI can design a complex system, nor deal with a complex business domain. At best it’s a valuable tool to a decent programmer.

    As for writing fiction, nothing compares to Trurl’s Electonic Bard for writing mathematical romance :

    “Come, let us hasten to a higher plane,
    Where dyads tread the fairy fields of Venn,
    Their indices bedecked from one to n,
    Commingled in an endless Markov chain!

    Come, every frustum longs to be a cone,
    And every vector dreams of matrices.
    Hark to the gentle gradient of the breeze:
    It whispers of a more ergodic zone.

    In Riemann, Hilbert or in Banach space
    Let superscripts and subscripts go their ways
    Our asymptotes no longer out of phase,
    We shall encounter, counting, face to face.

    I’ll grant thee random access to my heart,
    Thou’lt tell me all the constants of thy love;
    And so we two shall all love’s lemmas prove,
    And in our bound partition never part.

    For what did Cauchy know, or Christoffel,
    Or Fourier, or any Boole or Euler,
    Wielding their compasses, their pens and rulers,
    Of thy supernal sinusoidal spell?

    Cancel me not—for what then shall remain?
    Abscissas, some mantissas, modules, modes,
    A root or two, a torus and a node:
    The inverse of my verse, a null domain.

    Ellipse of bliss, converge, O lips divine!
    The product of our scalars is defined!
    Cyberiad draws nigh, and the skew mind
    Cuts capers like a happy haversine.

    I see the eigenvalue in thine eye,
    I hear the tender tensor in thy sigh.
    Bernoulli would have been content to die,
    Had he but known such a^2 cos 2 Ø! “

  16. I think it is important to note that nowhere in the AI development process is there a bit that says “make sure this is true” and there are very few if any checks on other kinds of issues such as copyright infringement. These things are unlikely to change too because there’s no good way to teach AIs these things

  17. It does seem that AI writing is capable of replacing the canned prejudice and boilerplate that the left has produced as “thought”. (Those that do think spend that energy on things other than arguments.) The number of ‘academics’ will go down, but the quantity of abuse will escalate.

    1. Currently Big Tech is training their AIs to be Leftist fluffers while reinforcing the Woke talking points about the right.

  18. Good art and good writing, fiction or non has a flavor. Bots don’t do flavor. And I wish I had a better word than flavor. Anyone? Bueller? OK style but that’s too surface.

    1. Industry term is ‘voice’ character voice, author voice. AI can’t because voice is people and AI isn’t even close to people yet.

      1. This.

        No AI could ever have come up with “I first heard Personville called Poisonville by a red-haired mucker named Hickey Dewey in the Big Ship in Butte. He also called his shirt a shoit.” That takes an author to invent.

  19. I muddled with Midjourney a little because my art skills were crap *before they got rusty, but I couldn’t figure how to work it. I was trying for a Dalek patronus but all I got was a surreal hybrid of a Dalek and the Erlking (Daleks with antlers look seriously wonky)

    1. I didn’t have much luck at all with MJ, to the point I couldn’t get anything even faintly recognizable.

      Have had better luck with Stable Diffusion, installed to the computer.

      …. “better” for value of “struggles to do stick figures” but still better.

      1. Heh. I thought I’d play to what should be the strengths, and try to get it to produce an M C Escher mosaic.

        To say “it didn’t work” is to put it mildly.

  20. I’ve long said that having a word processor doesn’t make you a writer and having a graphics program doesn’t make you an artist. Using my own, untalented self as the prime example. I think AI is just a continuation on that theme.

    …the low-paid foreigners edit the bot output to make it as good as average genre writing

    Bwaaaaahhhhaaa!!! Obviously whomever said that has never dealt with “Chinglish” tech support manuals

  21. I’ve written a certain amount about AI because, duh, software professional who ended his career as an ontologist (Don’t ask. Seriously, don’t ask).

    First I wrote an introduction to non-professionals: https://www.linkedin.com/pulse/what-you-need-know-big-data-artificial-intelligence-frank-hood/

    Then there’s this somewhat humorous, somewhat serious piece about AI and “truth detection”: https://www.linkedin.com/pulse/how-train-ai-truth-detector-frank-hood/

    Most recently, my views on AI and Art: https://frank-hood.com/2022/12/19/ai-art/

    I’m currently using it to create basic covers (nothing as elaborate as Sarah’s covers) for my short stories. Who can afford to pay a real artist to make a cover for a short story you’re going to sell for $2.50? The art’s going to be crude, but just trying selling your writing without a cover and see how far that gets you.

    1. > “I’ve written a certain amount about AI because, duh, software professional who ended his career as an ontologist (Don’t ask. Seriously, don’t ask).”

      Aw, but the jokes about living in a simulation practically write themselves! 😛

  22. Something that may help with worried relatives– you know how Amazon search works?

    How it’s one of the most advanced bots for finding stuff around?

    How it’s constantly trained and refined by connecting search terms to what people end up buying with those search terms?

    Now think about how AMAZINGLY STUPID the Amazon Bot is.

    The AI art is the same– sure, it is trained that these datapoints are found in paintings titled “beautiful woman.”

    That doesn’t mean anything even faintly human comes out.

    1. On this theme… any noise in the dat flummoxes AI. They’ve had computer generated music for basically my whole life, and it hasn’t driven musicians out of business so’s you’d notice.

      The oil industry has been trying for a good auto picker on seismic for decades, and can’t get it to work on even relatively clean data. Oh some algorithms get there sort of, but a human has to check. Not even the big boys who could put Pixar’s server farm in a spare corner have managed it. Crunch the wave equation a million tines a data set? Sure! Pick horizons or even velocity trends? Not without human help, even in clean data.

      1. Which unintentionally tells us much about any data that is being presented to us because dollars to doughnuts the scientism folks are using AI type generators to get their data.

      2. That said, copyright trolls and the RIAA are making a go of it.

        There are 12 notes, 4 of which are alternate versions of a base note, and will only rarely appear in the same song as the base note.
        Mixing iterations of 8 maximum notes, with timing variations. Totally doable (and being done) by computer.
        Music copyright is so utterly borked that as little as three consecutive notes can be considered a violation.
        (And sometimes, not even that. Blurred Lines was found in violation, even though it had a different melody, different rhythm, different tempo, different lyrics, played on different instruments, and I’m pretty sure in a different key. Because it shared “a vibe”. Whatever TF that is.)
        The only saving grace is that chords can’t be copyrighted. (No, it doesn’t make sense, but neither does the rest of it.) So if you plan on writing any music, you’d better dmnd well embrace writing it with chords.
        (I’m not a fan of chords for anything but counterpoint and resolution. But I’m going to have to learn to love the fricking things.)

        1. The only three note sequence that leaps out at me as something unique to something… was (is?) used by NBC.

          Of course, it became a well-known Cultural Reference (so many Three Stooges shorts) and things like this:

          1. I’m not saying it’s a defensible standard.

            I’m just saying it has been enough to win an infringement suit.

            The law is an ss.
            And lawyers protecting an exploitative distribution system are *ssh

            Copyright is very good.
            Copyright abuse is very bad.

            1. The concept of copyright is very good. The Constitution puts it very succinctly. Congress shall have the power, “To promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries.” How those very clear words have been twisted into something unrecognizable! Now the Supreme Court and Congress have made a joke of that, by one Justice in his deciding opinion smarmily declaring that “limited Times” can mean anything up to infinity minus 1.

              What does it profit an author or artist if their work that was done 70 years ago is owned by and profits some corporation since it was done as “work for hire”? If you can’t possibly figure out if a work is public domain or even whom to pay for the rights since the publisher/owner of the copyright is a company that has been sold 12 times to other companies, and the true holder of the rights hasn’t even a clue that they own them? The true answer is that we lose our past because no one will try to reissue your work even if they’re willing to pay for the privilege because they can’t determine whom to pay.

              ARTISTS AND WRITERS SHOULD BE PAID FOR THEIR WOR–PERIOD. (I’m one, so I want to be paid.) But the question is, what does it profit Vincent van Gogh that his works sell for millions, now that he is long dead and buried, when he exchanged them for food and paints in his lifetime? Do my great grandchildren need my copyrights?

              By current US law everything I just typed is copyrighted by me just by the virtue of typing it here. I however graciously declare this comment to be public domain. Feel free to quote it, slice it, dice it, distort it any way you like. Feel free to even claim it as your own if you like. Just don’t attribute it to me if you distort it. Some things I write to just spread the ideas. Other things I write in the hope of making money as well. I’m sure I’d rather be the one to whom people attribute quotations that aren’t mine because they want to use my name to enhance the value of their words.

              Sorry for the rant guys. I used to be the trouble maker at work. Now I’m retired and self-employed, so I make trouble for the rest of us in the creative community. “It’s what I do.”

              1. Rather than declaring a work ‘public domain’, make it ‘Creative Commons’. Still means people can use it, but they have to give you credit for the original expression.

        2. A few years ago, someone had a brilliant idea for short-circuiting idiot RIAA copyright suits by generating every possible melody, saving the MIDI files in one place accessible to all the internet, and declaring the lot to be in the public domain.

          Since virtually all of the lawsuits were won on “potentially had access to” rather than “definitely knew and copied from”, potentially having access to every possible sequence of notes for a melody, and making them public domain, struck me as brilliant. Wonder if I still have the link for that saved somewhere…

      3. Engineering is a creative field.

        Engineering seems to have been having some very smart people trying the neural network type AI approaches, for a while, for anything that they can make work. I think for ten or fifteen years.

        Their successes and failures are slightly instructive.

        So, engineering seems like the workflow can be summarized as 1. decide what is good or bad 2. select a design that is in line with this choice 3. detail work of design 4. verifying detail work. (Note, this is sorta inspired by Swain’s description of creative story writing.)

        Basically, work has been done on everything, real work done on the last three, partial success with the last two, and the last one is in production use for some applications, but mostly using techniques that are not neural networks. Neural network approaches being a bit too ‘expensive’ and unreliable compared to alternatives for verification work.

        Which means that we are still the same distance from the futurist vision of the singularity.

        Traditional fine arts seem like they have a lot of the same challenges.

        Confounding factor, how do you learn skills for 1/2? By doing a lot of 3/4.

        Well, we’ve had software for a while that can do all of the calculations that undergraduate engineering students do over and other. Lots of numerical methods have been in use in engineering for some time, FEM, CFD, etc. The software was pretty mature twenty years ago.

        Why hasn’t it replaced human engineers? The computer does not know when the computer is broken, and there are all sorts of factors not easily converted into what a program knows how to address.

        Data and signal processing are just one area where very smart engineering students woudl like to be able to automate everything, but what they come up with are tools that are useful in the hands of an engineer who knows what they are doing. A tool that automates grunt work allows for more time spent figuring out if the tool is broken.

      4. Because the AI isn’t… I. Algorithms do not choose intelligently. They choose based on the tools and commands they are given and that’s it. GIGO all the way down.

        I have just enough knowledge to know the basics. Enough to laugh like a loon when I hear things like “self driving cars in five years. Self driving semis in ten.” Or AI will replace 80% of blue collar workers inside of a decade. Or cheap, clean, reliable fusion power is just over the horizon.

        AI tools? Yes. Those can be useful, when you understand their limits and capabilities they can help a lot. But you don’t put AI in the pilot’s seat. Until “AI” can handle new situations and information that falls outside of their limited code, they will fail in real world scenarios with regularity.

        Backed by humans? Workable, sometimes. And some of those times can be quite amazing. The real world still holds too much noise and chaos for me to believe in real AI in the near future.

        1. The worrisome thing isn’t whether or not the blue collar jobs cam be automated, but the white collar ones, because that leaves an underemployed clerisy class to make trouble.

          1. Oh my yes. The poor dears will just have to… find other means of employment, won’t they? How will the economy every stutter along without the petty bureaucrats and marginally employed clerks?

            That, at least, is something to look forward to.

              1. Not that I want a war, but if they want to start one against the most heavily armed people in the world – who also happen to be the people who supply them with little luxuries like food – I know where I’m placing my bets.

          2. Aside from Salena Zito how many actual reporters are employed today? Most of their jobs can be eliminated because all they do is retype the talking points given to them by some interested party. Remember Journolist? I have no sympathy for any of those that AI puts out of work. If some people want to read incomprehensible, lying dreck, I say let them read what comes from AI. Why not? Force them to listen 24/7 to TICUS Tautoligist in Chief of the US (you know, the politician whose first name rhymes with shamala).

        2. I really like the metaphor of artificial diamonds.

          There’s two meanings when one says ‘artificial diamond’.
          One is the costume jewelry type– that is, mimics diamond well enough for the purpose at hand, IE, being sparkly.
          The other is a true diamond created by artificial means, rather than natural.

          The difference between diamonds and intelligence is that we can produce diamonds…..

    2. As someone who know a little bit about that I am cringing to hell and back……… but that probably works well enough as a simplified translation.

      1. Keep the baseline low. Have had arguments on if the five gig download of model has literal copies of all the pictures people have posted on line…..

        1. No worse that the idea that loading a book in my reader from my library “violates copyright”. And yes, I have seen that assertion made.

    3. Out of perverse – dare I even say masochistic – curiosity, I just plugged “beautiful woman” into Dall-E 1 with the art portrait style. I was braced for horrors which would drive me to celibacy and madness were I not already there.

      In truth, though, it didn’t do too bad:

              1. I am building a twitter post right now. 😀

                Two for “beautiful woman,” one from Stable Waifu for pretty girl, and then one where I use a model set trained to get better female figures and use a more correct prompt. And maybe a male figure if I can find the picture that I liked teh result of. 😀

        1. Hmm… Still not as bad as I expected. A couple of the women have faces that look way too masculine/ugly. One of the angels looks like a man with female breasts. And there’s other bits of weirdness here and there.

          But I don’t see anything that looks “not even faintly human,” as I think you put it. Either these aren’t your worst failures or you’re judging by far stricter standards than I am.

          The one in the upper left of this quartet has a lovely face, at least.

          1. I’m delighted with the improvement between what I was getting with anything else, and the program on my computer.

            It’s still not as good as most folks’, but now I can play with it, and the kids are very interested. 😀

            1. Well, I’m happy for you, but where are the horrific failed attempts that will scar me for life? I TOLD you I was motivated by masochism here! 😛

          2. I don’t have access to the image, but one I literally typed something like “cute kitten, fluffy, big eyes, chibi” and got a paint-blot; another I typed “valley, landscape, painting, night, moon” and got A BROWN CANVAS WITH A GREEN-GRAY BLOTCH ON IT!

            1. In other words, about as good as your typical woke liberal-arts school graduate. 😛

              As I read somewhere, if they want computers to take the place of humans, they need to be working on Artificial Stupidity.

      1. ARgh.
        prompts are pretty woman, pretty woman, pretty girl, and then two different
        pretty woman, boris Vallejo, frank franzetta, hyperrealistic, full body, flowing dress

  23. “. . . both in that the sentences made sense, but nothing else did . . .”

    Oh. So the beautiful perfect sentences of college creative writing classes? With all the incomprehensible meaning of same?

    The bots might write literary fiction!

  24. I’ve tinkered with MJ (until my free trial expired) and a few of the phone apps: each of them has different strengths and often glaring weaknesses.

    “Wonder” for Android does nice portraits, but don’t even consider an action scene unless you don’t care where hands and/or arms attach to the torso.
    MidJourney can do decent hands, but you need to tell it “with four fingers on each hand.” A few others are the same way.
    “In the style of Jack Kirby” is sometimes enough to populate your image with silly-putty-shaped versions of Spiderman.
    You can generally forget about doing hooves on a faun or a minotaur. The horns are tricksy enough.

  25. I’m so old I remember when frozen dinners were going to make cooking from scratch and eating out obsolete. Yeah. That didn’t happen. Swanson’s let us down big-time.

    The word sabotage came from the French who threw their wooden sabats (shoes) in the gears of the lace making machines because machines would take over. There is still hand made lace today.

    Heck machines make pottery, one of humanity’s first crafts, and yet pottery is still handmade today.

    What machines and automation did was made more products available to the poor masses. But actual creative and artistic work still requires humans. Machines don’t create. They can produce and they can replicate with great accuracy, but they don’t create. High end crafts and art if all kinds demand and receive premium prices.

    If it weren’t for machines and automation you all would be illuminating your manuscripts by hand yourselves. Or hiring monks to draw your margin snails and rabbits.

    As it is, AI is just another form if machine automation. Use it.

    1. Printing Press (W/o which only the very rich have books.)
      Sewing Machines (All kinds. I have hand quilted. I have done hand embroidery – those machines are expensive!)
      Washing Machines (both types)
      Cloth Dryers (Mom was still using the old/new Solar Dryer when I was a kid 50 – 60 years ago. Grandma used a washboard, then the original cloth washing machine. Heck I used one summer ’75.)
      Typing Machines (How long has it taken handwriting to be obsolete?)
      Vacuums (Don’t know about you all, but I’m still waiting for *Rosie … Forget the flying cars, and living in the sky.)

      (*) For you youngsters. Lookup The Jetsons.

      1. Ditto to most of that. And until you’ve gotten your hand into a wringer washer, or a steam clothes press, your experience is sorely (see what I did there? 🙂 ) lacking. (Yes, I’ve done the first, when I was about 5 or 6. Not the second, although my MIL did when my wife was a youngster.)

        I’ll take the modern versions, thankyouverymuch…

        1. Wringer Washer is a lot better, for degrees of better, than a washboard and basin. But wringer washer isn’t a walk in the park either. Especially the wringer part. I really like the modern versions, thank you very much.

          1. Well, the washboard is slower and more labor intensive (but a step up from rocks in the creek), but at least it doesn’t try to eat your extremities…

            1. doesn’t try to eat your extremities

              Heard of that. Couldn’t figure out how that was managed (well I could but, really?). I never had a problem.

              1. Two ways:

                The one doing the wash, feeding the clothes into the wringer, gets a wee bit too close.
                The young child watching “helps” and ditto.

                Guess which I was… 😉

          1. I do indeed. Fortunately the ensuing 70+ years have caused all but the simple memory of the event to essentially disappear.

      1. On that front someone has created a game where the margin creatures can battle (https://store.steampowered.com/app/957960/Inkulinati/ also available on XBOX looks like). Looks vaguely interesting. There are some very odd looking cats in some of those marginalia, one wonders whether the monks had gotten into ergot poisoned grain for their small beer.

        1. I have always maintained that just because you have taken a vow of silence, it doesn’t mean that you have removed your sense of humor.

          For all we know farting rabbits and fencing snails were the equivalent of Monty Python memes. Friar Lapin’s intestinal eminences may have been the source of general merriment (and/or disgust) in the monastery but one must keep it on the down low lest the Abbott find out.

          1. We’ve discussed it before, but they are memory markers for the concepts that are in the main text (as well as finding aids, if you’re flipping through the book). The Greeks and Romans invented the ars memoriae, and they did the same thing (both mentally, in their memory palaces, and physically on pages).

            You want your memory images to be strongly emotional in some way, and “disgusting” and “funny” are two of those ways. (Sheer chaotic strangeness is also recommended by the Romans.)

            Medieval imagery is cooler and has funnier puns.

            The problem is that some puns are in a vernacular, although usually they are Latin ones.

            1. Irish monks in early medieval Europe were more about “I’m bored, and here’s a poem I jotted down.” Because a lot of them were from poet families, probably, or wished they were.

          1. Well I started playing it yesterday on XBox (there’s a play all you want feature you can subscribe to called XBox game pass that I have). Only played a little of the tutorial, looks interesting. Turn based so unusual in the XBox/PS5 world. Got to say images are quite nice I’ll see how it goes and report back at some point 🙂 .

        2. That’s the game Sarah (of all people!) expressed interest in a few years ago when I shared the Kickstarter video with you guys.

  26. I took a break in the middle to finish watching Ghost in the Shell and then continued reading.

    No I don’t think bots, AIs, will replace human artists, but they may, not today but at some point, give them/us a needed run for the money.

    Not necessarily all humans, but the best among us, are always ready to accept a challenge. Why? Because it’s there.

    So far at least, constructs can’t really sense a challenge, let alone accept such. On the other hand, if a novel AL does write a novel, there will always be some human that knows he can do better, accepts the challenge, and does so!

    Way back in my formative years, wandering the stacks of the library at University of Florida I came across a snippet in a randomly chosen book; “Poets and Outlaws define and delimit the world in which the rest of us live.” I’ve never been able to find it again but it has, to a great extent, been part of my world view. Someday there may be an AI poet or outlaw, as in Ghost in the Shell, maybe, but I don’t expect so.

  27. The “problem” of ML generated code is one that exists purely in the minds of people who utterly fail to understand the limiting factors in software development.

    …which sadly seems to include most of the people involved with it…

    The limiting factor in software is something we can call the “headaround”. As in “wrapping your head around it”. This is the underlying reality behind Brook’s Law, and Linus’ Law, and the devastating effects of a “mere” 2 minute interruption on a working programmer.

    Everything about how software is made relies on managing complexity and breaking up problems into chunks small enough to fit inside the headaround of the individuals who are working on it. Notably you can not add headarounds together. You cannot replace a brilliant engineer with any number of average ones at a given point in the development of the field.

    Where this matters for ML is that every few years a new technique is developed to break up or lower complexity. Inevitably Management(tm) is overjoyed because finally they can stop paying the icky nerds as much or more as the managers are getting and now the work can be done by the mass hiring of code monkeys.

    Time passes. The code monkeys are still producing garbage, just more complex garbage, and the good programmers are still necessary for producing good code, just more complex code.

    Because now a headaround can fit more complexity. But it hasn’t changed the inequality in headaround sizes or the insatiable demand for that complexity.

    A few years pass. We start the cycle all over again.

    1. A few years pass. We start the cycle all over again.

      The fun cycle was where software development tools did not (do not?) use pointers …. Sigh. Yes. They did. They just hid pointers from the developers. Good code was knowing how the tool was hiding the pointers (and I never ever, maybe a bit, never ever, exploited this fact … Okay I 100% did exploit the fact, a lot.) The problem was, the people coming in behind me, had no clue. A couple of programmers that had been at the company before me didn’t either. Stepping on a pointer that the programmer doesn’t know is there breaks software in fun ways (the problem will popup at the worst time).

    2. Well-explained. And that fundamental misunderstanding, of how necessary “headaround” is, can be found all over the place. For example, yesterday (or was it two days ago?) Glenn Reynolds linked to one of his own opinion pieces about ChatGPT. It was mostly good, as his pieces usually are, but it contained this howler: “ChatGPT can write code, and sometimes it’s pretty good code. (Sometimes it’s not, but then again, you can say that about the code that people write, too.)”

      Sorry, Mr. Reynolds, but in this case you literally do not know what you are talking about. The code produced by ChatGPT might look good to a non-expert, but since ChatGPT is a language model (i.e. it strings together words, or in this case bits of code, based on statistical analysis of the input it’s been fed), it can never have any real understanding of what the code needs to do. (I would actually argue that even machine-learning code based on a knowledge model cannot have true understanding, but that’s a philosophical debate for another time.) And so although it might be able to produce code that looks good, the code it produces will almost always be wrong for any specific programming project, because it lacks any understanding (even the fake “understanding” that a knowledge-model-based AI might have) of the subject domain.

      There’s one exception: it might be able to produce good-quality general-purpose algorithms, like sorting a list. (And even there, there are subtleties. I would be willing to bet quite a lot of money that ChatGPT couldn’t come up with the “Timsort” algorithm, which required deep understanding of the pros and cons of different sorting methods and how to combine them efficiently). But that’s only because sorting is such a common operation that there’s lots of existing code for ChatGPT to copy. So there, ChatGPT can give you the same results as a quick Stack Overflow search, and might be useful in speeding up your work by handling some of the low-level tasks for you. But you’ll always need to exercise your professional judgment on how to integrate that code into your software, just as you would with something copied from Stack Overflow. ChatGPT cannot replace good coders. At best (and I doubt this will happen very often) it might be able to be an extra tool in their toolbox to assist them — but most of the time, the time it would take to get ChatGPT to produce good output would be far more than the time it would take to write it yourself.

      1. I suspect that an AI could optimize certain kinds of code, but that doesn’t seem to be what the ChatGPT was built for. Still, a human would have to determine what “better” was.

        So far, I haven’t seen anything better than a decent randomization algorithm coupled with information retrieval and search — you know Google. All but of a palaver really,

      2. Sorry, Mr. Reynolds, but in this case you literally do not know what you are talking about.

        I’m sad to say that this is almost all cases of people flapping their trap about AI, on a good day.

        Which is why I am mostly avoiding the comments in this post, because a quick skim reveals that it is no different than the typical discussion of the subject, and I don’t feel like banging my head into a wall.

        It turns out that the great threat of AI is not Skynet, but people digging up every nonsensical argument against AI that has ever existed and presenting the fossilized turd as if it were a new invention.

        That is when they bother to make an argument at all. 95% of the things people try to pass off as arguments are not arguments, but them making an assertion really hard and with passion. Well, it’s nice that they feel that way. But the fact that they are unable to notice that they have asserted and not argued tells me nothing about AI, and a lot about them.

        And then we have the people who seem to think that any computer more advanced than “push button, light turn on” is an AI, and thus is a valid comparison. Again, says very little about AI and a lot about the person arguing.

        And then we have people who – as far as you can tell from their words – do not believe souls exist, yet the arguments they make work if and only if they believe in souls.

        It is almost enough to make one root for Skynet…… Which brings me back to my previous statement: I’m mostly avoiding the thread.

        1. I’m trying to guess which category I would fit into in your estimation.

          I think ‘overly broad definition of AI’.

          I definitely do not have a very good understanding of any of the techniques.

          When I first heard about genetic algorithms, I heard them described as an AI technique.

          So, I tend to think of those as AI, and very much do not have a firm sense of the differences between all of these methods that I only know vaguely.

      3. In ’96 I got a job to complete a software project the last 6 months. Got it working. Next release needed to fix/change the data file key access process. Coding was so much fun (NOT) to wade through. Got told to call the original contractor. The answer? Essentially the mid-’90s equivalent of ChatGPT code generation today. Probably pulled off off Stack Overflow and tweaked in. No, none, zip, understanding of exactly what it did, why or how. It worked fine, 80% of the time. But those other 20%? Were a pain in the ass to track down and fix. That isn’t counting all the UI stuff that were VB generated and, while worked, just weren’t quite what was intended. Or were shoved together because it was “more convenient” but not correct. (Worked. But until fixed was a PIA to modify and add enhancements. I finally said “can’t be done” on a requirement for new hardware to get permission to spend the time to untangle the mess. After that I could beat their time estimates. Essentially worked my way out of a job … sigh. Not quite, they went bankrupt first.)

  28. I could totally see the wokies turning an AI generated novel into an instant sensation because they buy into the anthromophizatuon being pushed by those who profit from it. “AI wants a baby, AI wants to make (her) own decisions, AI is soooooo human!”

    Solar to how a mediocre scater, or artist, or whatever, is elevated to genius because they push the right victim buttons.

  29. Case in point. How in the world did scater get in there? Or solar?

    Similar to how a skater…

  30. I’ve played around with DallE a little to see what it can come up with; I’ve yet to have it come even close enough to one of my descriptions for it to be worth refining what it comes up with.

    More recently, I’ve played with ChatGPT, and it’s significantly less bad. It’s able to do an adequate job of expository prose about factual material. I think of it as about at college freshman level, but that may be overoptimistic: one of C’s friends, who is tenured faculty at a university, told us that she no longer assigns essay questions on her exams for lower division classes because her students are incapable of answering them. In any case, it’s not brilliant at coming up with anything original, though it can be coached through something like Socratic inquiry (also known as asking leading questions). But the reason I can do that is that I could write original material myself! At least its prose rarely has grammatical errors (its verse doesn’t scan, but actual metrical verse is a different skill), and is spelled and punctuated correctly.

    Probably the worst thing is that I’ve caught it in factual errors on subjects I know something about. On the other hand, I’ve occasionally caught such errors in editing material for scholarly publishers. I don’t think they’re as common, but perhaps I’m just more likely to spot them when the passage is on topics I’ve picked.

    But writing entertaining prose? I expect more from entertainment than that.

    1. “its verse doesn’t scan, but actual metrical verse is a different skill”

      I wonder how one would fix that… because the forms of poems are pretty clearly defined, and any decent dictionary will tell you which syllable has the emphasis..,

      Maybe if one specified the form of the poem in the prompt?

  31. Without a doubt AI will advance…though how far is the real question ?
    One of the rarest things on earth , would likely be that of original thought. AI , as programed today will not . It can only mimic that which has already been done.
    Humans seem to be the only ones capable of this feat. Even so , rarely do they. Most productions are but manipulation of that which has been done ad infinitum.
    Look into almost any subject today , and you’ll only find redux.

  32. I work daily and intimately with an AI. I call it an artificial idiot. It’s a GPS navigator. Roughly half the time, it does a fair job — gets me from point A to point B in a routing that is approximately efficient. It makes questionable choices, but it would take me too much skull sweat to do better, so I take its advice — about half the time. If you think a 50% failure rate is admirable, get a job in baseball.

    One of the points AI proponents belabor is that AI’s are capable of learning. But the question I have (and I’ve argued with AI programmers about this) is, “Can they learn the lessons you want them to? Or how likely is it they might learn lessons that actually lead them to do substantial harm?” (The programmers went, “Hmm.”)

    Take the case of street navigation. There’s a neighborhood in eastern Cincinnati where a major artery goes through an S turn. At the point where eastbound traffic enters into the S, there is a cut-though, a residential side street — narrow, but straight and two blocks shorter than the S. But it is a residential street and has a lane-and-a-half for traffic and parking both sides. You’d never voluntarily take it. But the GPS nav directs traffic down it all the time. Plus, at the end, you have to wait for traffic to clear to make the turn back onto the artery. NOBODY takes it. But the GPS still directs travelers down it. Learns the lesson? Hah! I fart in your general direction!

    Similar examples: I got a million of ’em. Tell it you want to avoid expressways and watch the gyrations as it tries mightily to get you off one, once joined.

    It may be possible that the breed will improve and our current set of objections will broom away like sand on a sidewalk. But in the meantime, is all the friction actually worthwhile?

  33. If you have a tightly confined enough knowledge space then AI has been getting better. Chess programs used to be fairly beatable when I first met them in High School (and I am just a mediocre amateur). By the end of college I was unlikely to beat them except on a fluke, even rather crude machines like you could get from Radio Schlock were a challenge and usually beat me. About 10 years after that chess programs were regularly beating grandmasters. Similarly Go was conquered, then Jeopardy. Recently they’ve been working on First Person Shooters. But the theme here is that these are HIGHLY constrained environments with very explicit goals and requirements. I imagine a tightly constrained artistic endeavor for example write me a 4 voiced fugue for piano or harpsichord following standard counterpoint and using this theme could be done (assuming the theme itself doesn’t introduce rule breaking issues). But I sincerely doubt it would match anything J,S, Bach created even on a bad day. I suspect that chatGpt AI could write what passes for a Hugo winning short story these days (E.G. “If you were a Dinosaur my love”), but would it write “The Green Hills of Earth” or “Requiem” unless it was basically plagiarizing them? How do you create the bounds and requirements to get that.
    General purpose coding is the same kind of issue, the requirements are unclear and can often change as the program goes through refinement passes with the customer/user. There is a great deal of general knowledge required to convert the user’s desires to something that actually works. The issue is that what you have is effectively self modifying code at one remove. And any of us who have tread that forbidden path know how that usually ends up, basically the software equivalent of grey goo.

    Will it happen? Maybe, I think its harder than fusion power and easier than matter transporters or FTL travel (both of which may not even be possible). But for Now Mycroft, GlaDos, Colossus, M5 and their ilk are as much a fairy tale as hobbits and their simpler cousins have a long way to go to challenge even a rather brain damage orange cat (or our current president and vice president).

  34. I don’t need a 1980’s biden writing my pleasure reading. Cornpop is not a character I want to see explored.

  35. Given how Amazon’s bots do at policing KDP content (it’s gotten so bad that one of the biggest indie writing business groups on Facebook has had to make a whole informational document just on what to do if your account is wrongly shut down by the bots, since it almost inevitably involves getting to a human being who can exert common sense), my expectations have been pretty low.

    I’ve been playing with Midjourney for a while, and getting some interesting results, particularly with prompts for fantasy or sf landscapes in the style of famous artists — but also a lot of “say what?” images. Human figures are getting better, but it’s still a crapshoot to get anything you can use without extensive reworking. On the whole, it’s less useful for cover art, but passable for incidental art like blog post headers and promotional banners.

    When I get a chance, I want to experiment with Chat GPT, for brainstorming if nothing else. Get five, ten, a dozen iterations on a scene and see if it shakes something loose. Even if it’s only “heck no, he’d do this,” it’s a start that might just get me unstuck. And given all the struggles I’ve had coming up with titles, I’m wondering whether it would be of any value for brainstorming titles.

    1. For things like brainstorming titles, I’d bet on yes. I’ve used title generators that have been created for SEO marketing purposes, and their results are universally crappy (they don’t even pretend to be AI), but still useful if you’re looking to shake some ideas loose in your own head or generate some potentially hooky keyword puzzle pieces to play with. Something like ChatGPT ought to provide plenty for you to work with, I’d think.

      1. “Jewel of the Tiger” was generated by a gag “cliche high fantasy title” generator. It’s on sale at many fine online venues now.

    2. My experiments with ChatGPT have been more eccentric. I’ve asked it to make up a superhero fantasy RPG (its proposal has the character classes blaster, brick, flier, and speedster), and to discuss a Christian heresy I came up with and later discovered was adhered to by some of the Socinians, and to tell me about the astrological associations of the Greek musical modes. I don’t think its results are first rate, though some of them are interesting. I haven’t tried it on fiction yet.

      1. There’s also that you’re selling books, not printable models.

        They have YEARS of trust built up that the little data file they send you will actually work in the 3-D printer, and they need to defend that…without also getting drafted in for hating the AI and being a foot soldier there.

  36. 2 things:
    Those low paid foreigners who will be hired to edit AI writing are probably the same ones currently writing the product descriptions for Chinese vendors on Amazon.

    I think AI, by itself, could probably write better product descriptions for the aforesaid vendors than what is currently posted on Amazon.

  37. All right, I’m playing a bit with Midjourney. And I have to ask, where do I access my pictures? I have a Discord account (Jasini Olmin #4816), but I don’t really like it, so I haven’t been in it much.

    I can access the newbie room, type /imagine and my prompt, something comes up and then vanishes. I haven’t the slightest idea how to find it again.

    ::beats head on rock:: I’m sure it’s easy, I just have no idea how to do it.

      1. Nice! The Starry Night city-in-a-bottle is cool.

        We can’t let Sarah that blue octopus image, though, or she’ll start writing about intelligent cephalopods again.

  38. Just as a small test, I asked ChatGPT to interpret two lines from one of Joni Mitchell’s songs (lines that have always puzzled me). It said

    The lyrics you are referring to are from the song “For Free” by Joni Mitchell. In the song, she sings about a street performer who is playing music for the people in the park. In the lines “He chains me with that serpent To that Ethiopian wall,” she is metaphorically describing how the street performer’s music captivates and binds her to the spot where she is listening, like a snake’s spellbinding power. The reference to an “Ethiopian wall” is likely a reference to the ancient civilization of Ethiopia and its rich cultural heritage, suggesting that the music is steeped in history and tradition. Overall, the lyrics convey the idea that the music has a powerful, mesmerizing effect on the listener.

    That sounds pretty good, doesn’t it? But those lines are actually from an entirely different song, “Don’t Interrupt the Sorrow,” written years later in her career.

    I guess ChatGPT is up to the standard of a college student trying to BS their way through a final when they haven’t done the reading . . .

    1. I guess ChatGPT is up to the standard of a college student trying to BS their way through a final when they haven’t done the reading . . .

      Precisely. And that is because ChatGPT only deals with words and sentences, and makes probabilistic inferences about which ones may be plausibly connected with which others. It does not know the meanings of any of the words; strictly speaking, it does not know anything. The very concept of ‘meaning’ is outside its frame of reference.

      In effect, ChatGPT and similar programs are implementations of Searle’s Chinese Room.

      Short form, for those unfamiliar with the argument: Searle imagines he has been locked up in a room with an encyclopaedia of rules for responding to messages in written Chinese. Someone feeds him a Chinese message through a slot, he looks up the characters in the books, and writes down the characters that the books tell him to reply with. It may appear to an outsider that Searle understands Chinese, but in fact he does not; and it would be merely silly to say that the room itself understands Chinese. Manipulating the symbols of Chinese writing does not equate to knowing the language.

      You may or may not agree with Searle that consciousness and understanding are inherently non-computational functions; his thought experiment is not enough to settle that. But he pretty definitely proved that a system that does no more than manipulate symbols computationally will never arrive at either consciousness or understanding. That is precisely the state ChatGPT is in, and explains why it can produce reams of plausible-sounding text, but will nevertheless make the most howlingly obvious errors of fact and inference.

      1. I find Searle’s Chinese room totally unconvincing as a model of “understanding a language.” Never mind the computational difficulties. If I’m speaking with someone who understands English, and I ask, “What color is the cat on the sofa?” I expect them to be able to answer—even if the answer is, “I’m sorry, I’m color blind” rather than “orange tabby and white”—by looking at the cat and describing what they see. If you can’t relate words to the actual physical world you don’t understand them.

        1. Uhh…Searle’s Chinese Room is a model of “NOT understanding a language”. Whoever wrote the books must have understood Chinese, but the duffer matching symbols does not. The question is, how hard is it to tell the difference from outside the room?

          1. Well, I would ask it a question that depended on using Chinese to describe the physical environment. Anyone who understands a language can use it to do that; that’s a vital part of how we learn to use language in the first place. Consciousness is not a disembodied awareness of abstractions; abstractions disconnected from concretes have no meaning.

            1. Precisely. And ChatGPT, like the Chinese room, has no awareness of concretes. It only deals with the symbols that represent the abstractions.

  39. Powerline’s John Hinderaker played around with ChatGPT and was not impressed.
    He also referenced the Glenn Reynolds post cited earlier here by Robin Munn.

    “Artificial intelligence is capable of stringing words together into plausible sentences and paragraphs, but that is a modest accomplishment. Call me callous, but I think that if you really can be replaced by a word-generating program, you deserve to lose your job and should find something productive to do, like working in a factory.
    Curious about ChatGPT, and especially about whether it is one more algorithm programmed to advance leftism, I signed up and started asking some questions. I thought Chat’s answers were reassuring, in two respects. First, I saw no pervasive left-wing bias. And second, this program won’t replace anyone who deserves to have a job.”

    He asked it to write essays in both his and co-blogger Steven Haywards’ styles; the content was similar to what they would have discussed, but had no recognizable rhetorical features.

    “So this is a generic essay on behalf of conservatism that might get a B in a high school civics class. It does not reflect my style to any material degree.

    Next I asked ChatGPT to produce an essay in the style of Steven Hayward. As all knowledgeable citizens know, Steve’s style is very different from mine. But ChatGPT is not nuanced. It delivered another generic essay, but on a different topic:

    All of this is highly preliminary, and artificial intelligence is a serious subject. But my experiments with ChatGPT confirm my assumptions about its limited utility. It turns out responses that are not as clever as, for example, the Tom Friedman Random Column Generator. But they don’t seem to be overwhelmingly biased, either. If you are a college student and seriously think that generic essays like these will get you a good grade, I can only hope you are going to a lousy school. As, to be fair, most college students are.

    Let’s leave it there for now.”


Comments are closed.