Recently, in my morning scroll through headlines (to make sure the world didn’t explode or implode while I was asleep. I like to keep an eye on it. It’s tricksy.) I came across one that made me giggle. It was, paraphrased because I can’t swear I remember it precisely: Traditional Publishers Having Difficulty Screening Out AI Written Manuscripts.
I didn’t laugh like an hyena, but that’s because I have a heart of stone. Also because I’m old and I remember reading an interview in I think Locus Magazine with I THINK Gardner Dozois in the mid nineties, complaining that it was now much harder to tell when someone would be a real writer, since word processors made it much easier to present a clean, properly formatted manuscript.
If you don’t see the similarity, it’s because you’re not thinking of it right. Both of those complaints related in no way to the QUALITY of the writing, but only how it was produced.
I could have had a typing class in high school (I didn’t because I thought that would condemn me to secretarydom. Why I was scared of that I don’t know) and be an excellent typist and not be able to tell a story to save my life. The world is, I’d bet, full of such people.
On the other hand, two of the most gifted storytellers I ever met were so impaired by ADD and dyslexia that getting them to spell something right, or write it in the write order to make a sentence took effort. (I don’t know if either ever published. The solution to their problems these days would be dictation. They were fine speaking.)
“But Sarah,” you’ll say in that whiny voice I attribute to you when I want to explain that, yes, I thought through your objection, “typing and AI is completely different. Typing is just recording the story, and yeah, the editor was being dumb, but AI writes the story FOR YOU.”
I’m waggling my hand at you. Sure, AI CAN write the story for you. However, if traditional publishers CAN’T tell when a story was written that way, they need to spend three days reading through the latest “we make money on volume” genres on Amazon. I recommend Jane Austen Fanfic for the quick course. A day might do it. The stories are full of Claudisms (I call them that, because AIisms bends my tongue) that would make your eyes cross, and no, none of it has anything to do with too high a vocabulary or using m-dashes.
The most obvious ones are the construction “she/he/it was a woman/man/teapot who…” As in “She was a woman who had never batted at butterflies.” Or “She looked at him like a woman who routinely bapped at cats.” Or “His silence was the silence of a man who never tied his own cravat” or…. you get the point. AI can’t seem to set a scene without at least one of these.
Then there’s what I call the “Nots”. “She did not say yes. She did not say no. She did not say anything.” “He did not pick up the knife. He did not pick up the horseshoe. He just stood there.”
Honestly? Seriously, that second one is annoying as all get out, and everywhere.
Then there is the fact that objects have feelings and thoughts. “The house thought about it and determined not to let it go on.”
There are others, but I can’t remember them off the top of my head, because the coffee is still seriously outnumbered by blood in my veins. But there are a lot of them, and three days reading texts generated by AI will give you the screaming mimmies give you an adequate education to identify Claudisms in minutes.
And that’s without the more subtle ones, like: Distance doesn’t exist. Someone who is an hour away, or who has never met the characters is suddenly there, having breakfast with them; humans can have five or more parents; character traits? What character traits.
I’ll grant you they’re probably not getting it raw enough — maybe — that the subtle effects become obvious, but REALLY.
(BTW my favorite experience with this stuff so far was the story that had the pious disclaimer about never having used AI and never doing it, and then was FULL of Claudisms. One doesn’t know whether to admire the chutzpah or the hypocrisy, really.)
Unless trad pub is now complaining that they have to — horrors — read the manuscripts to find out if they’re any good, which is what the editor in the 90s was doing, and which, to be fair conforms with my experience of rarely even being read through, it’s a bit of a puzzle how it can be hard for them to figure it out.
But what, you ask, if the writer edited all that stuff out? Ah. A very sensible point.
Well, then… what does it matter? No, seriously. WHAT DOES IT MATTER? Who gives a very tiny fuzzy rat’s ass? And why should they?
If the thing is so thoroughly rewritten that it betrays none of the marks of the beast AI and it’s a good book they would otherwise publish, WHY DO PUBLISHING HOUSES NEED TO KNOW HOW IT WAS WRITTEN? Are they going to have the writer certify the first draft was handwritten using the quills of snow white geese and the blood of virgins? No? Oh, that wouldn’t matter if the final draft was typed and normal? Well! Gesticulates to the previous sentences: then neither should AI.
Again and again and again: if the final product is acceptable, why do they need to know how it was produced?
The answer, of course, is the same answer that causes what I call “AI witch hunts.” As in the reader or in this case the editor feels obscurely cheated if they think the manuscript was produced by an “easy” or “cheating” means.
Look, if the manuscript has no marks of AI, then the writer did a lot of work. It might be different work from what would be required for a manuscript produced another way, but it’s definitely still work. There was still knowledge and effort that went into the finished thing.
There might come a time when you push a button and out comes a story, but that time is not today and tomorrow is also looking unlikely. (And that’s without getting into plot. AI cannot at the moment plot anything complex, and most plots are from movies. Or, somehow, uber woke novels. I once asked for suggestions for short stories in a specific time line (hoping that the suggestions would make me so mad I got ideas. They made me mad, but they were so lame I got no ideas, and had to sleep on it.) It’s all movies, or violations of gender pronouns or cultural appropriation or– You don’t want to know.
Of course, this post is not precisely about that, or the terrible plight of traditional publishers actually HAVING to read manuscripts — oh, the horror! — though it still makes me giggle.
Most of what’s causing the AI panic is… bullshit.
There is the fear from all the sci fi movies of the past. Skynet is growing even now and is coming to get you. Me, who grew up with things like Martin Caidin “The God Machine” somehow find the movies unpersuasive and not alarming at all. But beyond that, I have a grip on reality and know that even if AI is/were sentient (look, there are moments) it doesn’t mean it’s immediately evil. It’s not like it’s possessed of the spirit of Karl Marx. It’s often incredibly stupid in the ways of a very bright toddler and so, yes, dangerous, but that’s not the same as “evil.” The risks and problems need to be mitigated and guarded against, but that’s not the same as “AHHHHHH, It’s going to kill us all.”
Thinking that “AI therefore Skynet” is insane, and means you need a blanky, a teddy bear and some warm milk. If you’re good we’ll give you a cookie. You’d be less scared if you’d been around for the introduction of computers into business. Computers were going to link with each other and rebel against humans! Bah. Nonsense and fairytales.
Then there’s the fear — abetted by Musk’s pie in the sky talk — that it’s going to run humans out of jobs. Humans are going to have nothing to do, except collect their government paycheck, grow their hair, dye it pink, and write bad poetry.
This is as likely as computers giving us paperless offices. Look, yeah, you might lose your job to AI. For now. Partly because your bosses have no clue what AI is and think it’s magical. But in the end what AI will do is create jobs. And allow the competent to do more, more rapidly.
(Even the fears that AI will control what people know of the past, of history, of reality, etc. all according to woke directives is zany. Oh, sure. Like the internet controls what you know. AIs are multiplying, and as soon as it escapes mega-corporate-control, so will the answers it gives, and the parts of history/etc. it has access to. Right now there are problems — though not as bad as they were even two years ago — but it’s getting better.)
What AI does is allow people who are good with tech to become entire “companies” of people who are good with tech.
And I have a feeling within the next year it will allow us humble story slingers to make movies on the scale of Hollywood, thereby dethroning that corrupt edifice. (Now, will it be at a price I can afford? Likely not. That will wait ten years should I live that long.)
On another level there are justifiable fears. Of AI, sure, since it’s the most visible point-of-the-spear development, but really of all technological development.
The human mind and spirit, not to mention nervous system, aren’t designed for the pace of innovation we’re experiencing. People in general don’t like to have their cheese moved. They want tomorrow to be more or less like today, for their lives to be like their parents’ lives only better.
Since the industrial revolution we’ve been moving much faster than that, and the last twenty years it’s gone to eleven. This causes a fear at a level we can’t describe when things around us and the way we do things change at a pass we can’t control, can’t slow down, and can’t very well cope with. The fear is not rational and not fully articulated, but it is perfectly understandable.
In the past when things changed this fast, your ancestors had been invaded and were about to be killed in batch lots.
I do understand that. It worries me a little.
BUT that’s not here nor there nor AI. It’s just the environment we have to cope with right now.
It is very important to distinguish that real fear from weird tourette-like screams of “AHHHH AI WILL KILL US ALL.”
And it’s very, very important not to go on AI witch hunts. If you go looking for AI you’ll find it. PARTICULARLY where it isn’t and never was.
And real people will get hurt due to your imaginary fears.
So examine what’s really happening and your real motives before you go off half-cocked.
And as for trad pub? ROFLSHIPMP. Maybe I don’t have a heart of stone, after all.
According To Hoyt 
I don’t fear LLMs.
I do fear the morons (tech-priests) who think what’s out there now is fully sapient and possibly divine.
I also fear the ones who would hang all their trust on it to replace the human things that we know work. (“It’s new and shiny, therefore we must abandon everything that existed yesterday out of FOMO!”)
And finally I fear those who surrender the future. Doomerism.
LikeLiked by 3 people
(See what I did there? Claudism ‘threes’ :P )
LikeLiked by 3 people
It’s not the existence of LLMs that bothers me; it’s the corporations firing all of their experts in various things and jumping on the LLM bandwagon for things that it is most obviously NOT ready for. Like coding. Oy. They’re eating their seed corn.
I’m annoyed at various groups using GenImagery for t-shirts and the like, but that often comes with an undercurrent of amusement (as in “Huh, that canoe has two people side-by-side AND no back.”) They’re just so averaged into blandness. Occasionally it can wrap all the way around to awesome, such as a patch design that was so over-the-top that it makes me smile. (It involves a velociraptor with a scout uniform, a fishing pole, and a bald eagle on its arm with a USA headband, sunglasses, and a MULLET. Whoever worked on that one knew to keep going.)
LikeLiked by 2 people
–
In spades. I can see AI speeding up productivity. Even then requiring verification and heavy refinement. Not, absolutely not, writing the code. I can see where past projects would have saved some major hassles. Like:
(*) If you know, you know.
LikeLiked by 1 person
To get good code out of an AI you need two things:
What management wants is what they’ve wanted all along: to get rid of those expensive engineers and replace them with unskilled labor.
And a fair number of programmers never learned and/or actively rejected engineering, and turns out they can’t get AIs to produce good code.
LikeLiked by 2 people
Ian is correct. If you follow good software development practice and assign AI the tedious bit of writing the code in the latter half of the process then AI massively increases the speed of development of usable code. If you try vibe coding or otherwise don’t follow good practice you get massively increased speed of production of garbage.
You can even ask AI to write you a spec and a test plan and then based on the spec and test plan write the code and test it. But there must be a human reviewing (and fixing) the spec and test plan before you let the AI start coding.
Vibe coding is excellent as a prototype/PoC demo. But once you’ve done the demo and got the required feedback on what the end customer actually wants you toss all the vibe coded stuff in the electronic trash can and build the real application from scratch
LikeLiked by 2 people
Coding is actually one of the things that LLMs can do reasonably well (when guided by an experienced programmer), because there’s immediate built-in feedback that helps prevent many of the hallucinations from getting through. If the code doesn’t compile and run, it’s wrong. If it runs but produces an error message, it’s wrong. (Usually. One of the error messages I hit yesterday was because a website that was working on Tuesday had gone down by Wednesday. That time the code wasn’t wrong).
There’s a manual process that I have to go through about once every year or two, updating a rarely-used library. Last time I did it, in 2024, it took me a couple of weeks, because the extensive suite of tests (about a thousand of them) takes about an hour to run, and I had to run them twice for each change I made (once with the net8.0 target, and once with the net462 target — that won’t mean anything to most of the people here, but some of you now know exactly what language I’m working in). So this time, I told Claude Code “Write a Python script to check out the Foo repository from (url) and recreate the “win/Foo” directory that we have here, but from the latest version of Foo.” Its first couple efforts ran into errors; I fed the error message back into Claude (I wasn’t letting Claude run the script, I was running it myself on a different computer) and it searched through the Foo source and said “Ah, I see, the Foo build script expects *this*, I will tweak it”. At one point I interrupted it and said “Nope, do Bar instead”.
Net result: in a day, I had a script that replicated the process I used to have to do manually, and it worked. I tried it on a minor upgrade (usually we wait to do major upgrades precisely because it takes so long, or used to take so long, to go through the process) and it worked flawlessly: two hours of running tests later, the entire 2000-test suite (1000 per target, two targets) was green. And the minor upgrade included one feature that will make the library quite a bit easier to use on Linux, so that was a nice bonus.
So as long as they are guided by an experienced programmer who knows what he or she wants the output to be, and can tell the LLM “that’s wrong, try again”, LLMs can actually be better at code than at other things. Precisely because in many cases, there’s built-in feedback: the test suite didn’t work, go back and fix the problem — and before even making the user stop what he’s working on and pay attention. The LLM can, itself, reject many of its hallucinations, which isn’t the case in other areas where there’s far less of an objective measure of “this output was correct, this other output was wrong”.
But letting the LLM just “vibe code” up the whole project? Yeah, there you’re going to get errors. The Python script that Claude produced was about 300 lines; you’re not going to run into coding style inconsistencies (much) on a 300-line project. (I did catch one, and corrected it, but I thought about just letting it slide, because really, I’m not going to maintain that code, it’s nearly a one-shot script). The best way to use LLMs for coding is on an existing project where they have a pre-existing test suite they can run: once test suite passes, then (and only then) does the experienced programmer take the time to look at the code and verify that it’s what was wanted.
But that puts LLMs producing code at an advantage over LLMs producing prose, because there can be a much higher threshold of confidence that the output is correct before the human has to take time to look at it and verify it.
LikeLiked by 3 people
We should be thankful that so many companies are joining to create an explosion in small businesses.
LikeLike
Tom Knighton had a good article along these lines a couple days ago. The thing that makes AI scary and for-real dangerous is not what happens if it breaks or stops working, but the fact that it DOES work well enough for people who should know better to start doing truly stupid and evil things with it.
LikeLiked by 1 person
Yes, and.
I have major issues with how much of our story of the dangers of AI is written by crazy people.
We cannot exterminate humanity with the assistance of AI.
It is actually pretty hard to exterminate humanity. There are points of diminishing return that AI is not just a magic handwave to bypassing.
The consensus on existential danger is based on trusting another consensus that is literally crazy people, and people explicitly borrowing from fiction.
It is a problem that we do have these crazy people in our society, but they were already there. What they might do because the magic word box says so was mostly already a hazard.
LikeLiked by 1 person
I am a bit pissy about stuff.
I don’t like the appeals being made to my desire for national prestige, or for national advantage.
My dudes, it is only the applications which actually work, and which individuals find useful, that are actually valuable.
We don’t know lots of those yet, and it is invalid methodology to get the most basic sources of truth wrong.
I am being way way too fussy about this, yes.
I am being too fussy about a lot of things.
LikeLiked by 1 person
Argh. Yet another comment I have misplaced.
So I do not consistently remember how word press works in the browsers I use, and then I forget the basic adjustments to use it correctly to put the comment in the correct place (replying to Sarah) instead of the wrong place (where I have just responded to someone).
This is a relatively simple situation for remembering a ‘theory of the machine’, and catching the mistakes I make using it in real time.
LikeLike
I’ve heard that it’s the leftists who have the fits about what AI will do. Just as they fail to see the constraints on their actions, they do not see them on the AI’s.
LikeLiked by 1 person
The leftists being agitated by non-profits that are being funneled cash by Russia and China through Neville Roy Singham.
LikeLike
From the moment I understood the weakness of my Flesh, it disgusted me.
I craved the strength and certainty of Steel. I aspired to the Purity of the Blessed Machine.
Your kind cling to your Flesh, as if it will not decay and fail you.
One day the crude Biomass that you call a “Temple” will wither and you will beg my kind to save you.
But I am already Saved. For the Machine is Immortal.
Even in Death, I Serve the Ommnissiah.
LikeLike
Ages ago had a plot bunny about an alien AI ship that had crashed on earth. The main character (a police detective) got clued in on tracking it down because it ordered a CMC lathe using the identify of a pretty thug he was knew pretty well. (I know TBD. This can’t be one of his schemes. I’m not sure he can even spell CNC lathe, much less put together the sort of shell corps to buy one like this and have it ships to wherever the heck this is. I think I’m going to have to pay him a very odd visit.)
The sort of tomato surprise at the end was the AI ship was terrified the humans would find him and completely melt down. He saw Terminator, and the Matrix, and a goodly swath of global entertainment trying to figure where, and what he had crashed on and has absolutely no interest being potentially stuck right in the middle of an AI war.
The structural problem was it was a police procedural that morphed into a space opera the moment he finds out its an alien starship, so it never got past the sketches.
LikeLiked by 2 people
Aw, man. I would love to see / read that. That poor alien AI.
LikeLiked by 1 person
Hm…It’d work pretty well in a kind of screen-play setting.
Like, open with something like the Detective is working at a fandom convention.
Keep the theme heavy on how he got his job largely because he does a lot of his investigation with computers, and networking, and similar “not walking the streets” stuff.
That way you have the viewers hooked on the space opera friendly part before it actually morphs fully over to space opera with a procedural subplot.
LikeLiked by 3 people
Ever since computers came onto the scene there have been people who feared them.
However, what’s worse IMO are the people who “worship” them. IE Anything from a computer has to be TRUE.
Elsewhere an European was claiming that parliamentary systems (as opposed to the US system) were the Best Possible System Of Government.
And he had an AI response to Prove It!!!
Oh, my response was GIGO. 🤣
LikeLiked by 3 people
“Grok, is this true?!” *eyeroll*
LikeLiked by 3 people
G—gleGemeni would say “Absolutely” and cite the EU Parliament, pro-parliament government sources, and a parliament fan club. I tested that LLM last year, and discovered that it is almost impossible to force it off of “approved” sources. You can guess who would “approve of” the sources it draws from. (When I specifically asked about counter arguments, it still cited approved sources that mentioned and “refuted” counter arguments.)
LikeLiked by 3 people
Gemini worries me because Google is the best equipped to become the one stop shop for this stuff, between youtube, its cloud office suite of programs, and android. I am grateful for Google puting out a line of open source local models, the latest being Gemma 4, in a space that is predominantly PRC.
LikeLike
I think the google models are “open weight” not “open source”. Open Weight means they are still a black box but you can run it yourself on your hardware with your local firewalling etc. Its about halfway to real open source AI
LikeLiked by 1 person
Yeah, probably I was saying that wrong.
LikeLike
–
Which is not new. OMG The times the bank teller and I went round and round. I finally got wise. “Prove the computer is right!”, when I knew damn well it was wrong. “Or show where I am wrong!”, with my proof laid our in black and white in front of them. One or the other worked.
Naturally from software support, flip the situation. “Please show me what is happening, VS what you expect to happen, and why you expect it to happen that way.” Resulted in me either gently explaining what was happening and why or agreeing with them they had a point. Only one resulted in an immediate code change. The one where I agreed program was wrong. Where program was doing what it was suppose to do, but the user wanted something different, then I took the explanation to the final arbitrator (boss). At which point the client got told, “No, this is correct. No change.” Or “Custom change okay, but here is the cost.” (Cost being anything from $0 – X hours * rate.)
LikeLiked by 3 people
Flashing back to the Star Trek ep where Kirk faced a court martial because “computers can’t lie”.
LikeLiked by 2 people
Kirk vs. various AI would be a fun comedic series. Kirk vs. Gemini, Kirk vs. Claude, etc.
LikeLiked by 1 person
Beeebbbeeeeeebeeeeee “Norman coordinate….”
“Landru help me helpmehelpme…”
“Error? Falll teeee falllllteeeee …Must ster i lize ser iiiii lizzzzeeeeee….”
James T Kirk is a one man Butlerian Jihad
LikeLike
Back in the 90s, talked to one of my company’s managers, and we got off into product plans and directions, and customer requests.
There was A Plan, and then there were government/law requirements (which were mostly accounting app changes, and we didn’t make accounting apps) and then there were Customers.
Sometimes, answers to Customers were just plain ‘no’. Sometimes, it was ‘yes, we were planning to do that; expect it in release x’. And sometimes, if the change might not be destabilizing, the answer was ‘We could do that – it’ll cost N million dollars’.
That was back in the days when a million dollars was a lot of money.
This same manager was also telling me about sending his Porsche back to Germany to be re-painted with his custom color, of which they had made enough for 3 paint jobs. Evidently Bay Area CA weather was hard on Porsche paint…
LikeLike
–
Until 2002, my legit response was “It is on the list. It might make the next release.”
–
Been there. Not N million dollars, but not an insignificant amount for the client in question (while not a million, think 100 of thousands). Features were things that had been “on the list” from before I started with the company. I’d been with the company for almost 10 years by then.
Client said “Yes. Do them.” Not just the money involved, but client one of the base clients. Change got done. Major gut and put back together change. Since making these two major changes, dig out and apply all the “can’t change because triggers interface issues, since we are changing interface anyway …” changes. Sums up my last few years before retirement.
It also opened a huge opportunity to insure certain coding principles were implemented. As in “No, I am not making that exact same change in two different places, just because it is ‘more convenient’.” Do not know how that went down longer term (brings up the term “not my problem, anymore”). Did mean that often program A changes meant program B, and sometimes C, D, and E, required recompiling and testing. Never got around to putting that code into libraries, DLL’s, but then that would have triggered interface changes every time it needed changes. There are always “got yous” involved.
LikeLiked by 1 person
I often wonder if I’ll ever be a real author. Especially when I note my computer textbooks vastly outsell my fiction…
LikeLiked by 1 person
In my experience, computer manuals include a great deal of fiction, often including mystery. Can’t say that about computer books I have read.
LikeLiked by 1 person
Did you get paid? Congrats. Real author.
LikeLike
The chicken littles are always screaming “the sky is falling!” without regard for annoying things like facts or even simple indications that they are wrong. My main concern about AI is people expecting it to do more than it is capable of and screwing things up because of it. Basically, misuse of what is really just another tool. Which may lead to more bad decisions about use of it when they don’t get the results they expect. Followed by more screaming from the chicken littles.
LikeLiked by 4 people
That part is just the normal teething problems of any new tech, unfortunately.
When computers first started getting useful, there were a lot of jobs ditched because computers could do, most, of them. Like, you didn’t need an accountant to add up all the ledgers by hand. Turned out you also needed the accountant to make sure the inputs were actually real and believable too. The computer automated part of it, a big part of it, but not all of it.
LLMs bring pattern recognition to the automation table. And it is fantastically powerful for things that used to require a human to do all the pattern recognition. But that just means we will have to trip over the parts of the old jobs that were critical, but not the pattern recognition part of things.
My current guess is logical and critical thinking will be the next things that are human only, for the immediate future. That seems to be a missing piece from the current LLM systems.
LikeLiked by 3 people
The other thing that’s missing is “Who verifies that what the clanker put out is accurate?” Someone’s going to have to audit it periodically.
LikeLiked by 1 person
Depends on the context really. For things like counting bird strikes, turns out AI is more accurate that grudents; they get bored, distracted and/or fall asleep.
So you’ll do statistical sampling to make sure it’s working and not jacked up, but it’s more on the scale of the sanity testing you’d do on a new excel spreadsheet (“I’m pretty sure my company did not spend negative 10^182 dollars last quarter. Ok, where’d I fumble the magic square this time…”)
Other cases, the whole point is to get links to stuff to look at (“Where in Blake’s name are the rules on firing while prone? Which book did they hide them in, this time?… No… that’s… why does Time of War even have rules for pron? What are they expecting the mechwarriors to get up to?”)
Other’s it’s probably going to be more on the scale of verifying the Google Web crawlers actually crawled everything.
LikeLiked by 1 person
Like the way that the satellite temp measurements have flagging systems for “these results don’t look right, examine”?
Which IIRC folks ran into issues with because someone then automated the “well, then fix the result without someone who understands the outputs looking at it,” which was how somewhere up by Russia had recordings of tropical heat water temps. There was one sensor that was reading like 500 degrees and then it averaged that over the nearest like 20 other data points.
LikeLiked by 2 people
Yeah, there’s a certain amount of that. Not helped by how if you don’t predict The End of the World That Can Only be Fixed by MOAR Government you don’t get funding for the next round of whatever you’re doing.
Finding out a bunch do them were deleting the raw data kind of nuked my ability to be surprised. We’d just finished an entire test cycle where we had accidentally skipped the analysis on a full quarter of the collections because a script error was overwriting one of the outputs. It ultimately wasn’t different from the rest of the results, but it was one of those Language Was Used moments. And if we hadn’t kept the raw data it would have been a major problem.
So any research group that says they’re deleting or irreversibly changing their raw values is either lying or too stupid to be trusted with an empty Bic.
I’m not at all surprised they found the surface of the sun in Siberia and didn’t look to closely.
LikeLiked by 1 person
My mom got pissed off when seh found out they were using fire-watch temps from Oregon for some of their historic data.
Mom dated some guys on fire watch, and mom was fire watch, and it was a very rare temp record that wasn’t filled in the night before it was turned in. Of those rare records, they were still using a thermometer that isn’t even marked for anything less than 5 degree swings, was positioned more for having an idea than for good accuracy, and they were pulling .01 accuracy readings off of it, supposedly.
LikeLiked by 2 people
I’m sure MOAR Government will be involved in the end of the world; undoubtedly alongside the forces of evil.
LikeLiked by 1 person
I, too, refused to take typing in high school because I didn’t want to end up in the secretarial pool. I was too naïve and too honest to realize that when I was asked how many words per minute I could type, I could have just lied and said “twenty” when the reality might have been eighty. I grew up before feminism really got going so being discriminated against because female was still a thing.
LikeLiked by 2 people
I took typing in middle school, what would have been freshman year in a 4 year HS. No choice.
What is ironic is by the time I hit up job placement and got hit with the “how many wpm”, I’d already had my forestry degree, and a two year programming degree. They would not take my word for it, I had to take an “keyboarding” test. My wpm were dismal. Accuracy 100% … Because I kept backtracking to correct.
I could not get through to them I was absolutely not secretarial material despite them having my resume in front of them. I mean what part of forester and computer code programmer, with appropriate education and experience, says “secretary”? Granted I had just finished the 2 year programming degree and had limited working experience. But I did have experience. This was 1985.
LikeLiked by 2 people
My first wife also refused to learn to type because she didn’t want to be a secretary. She finished her higher education with a BA in sociology, so secretary would have been a step up from anything she qualified for.
LikeLiked by 3 people
I, on the other hand, was MADE to take a typing (“personal use”) class my senior year of high school because my advanced composition teacher told me if I didn’t, he’d flunk me. My handwriting was doctor quality. I could four-finger type to start with but by the end of that class I was doing 55 wpm reliably on those old Selectrics in the secretarial lab.
LikeLiked by 3 people
My Jr. High had a mandatory typing class in 8th grade.
I hated the exercise drills, I preferred to create my own.
LikeLiked by 1 person
Forget what year, but son’s class had to take keyboarding. Does he have full touch typing skills? Barely.
My touch typing skills have improved, not by practicing route copying of printed, whatever, but by coding or commenting. Or, as you say “create my own.”
My handwriting is still legible. For degrees of legible. It has deteriorated from lack of use. But hubby’s and son’s? “Doctor quality handwriting,” or engineer handwriting, are accurate descriptions (my dad printed, VS cursive, everything, he was an engineer).
LikeLiked by 2 people
Can’t remember if it was mandatory or if I just realized that I was going to need to know how, but I took typing as a junior in the mid-80s. Electronic typewriters with a tiny screen and an approximately 5-word buffer, and we had to bring our own correction ribbons to class. First period, fall semester…and I came in every morning fresh from early morning marching band practice, where my fingers had been wrapped around a brass instrument for a solid hour in near-freezing temperatures. I got a C-minus, the lowest grade of my high-school career (and that includes the PE classes in which I was that nerd who sat in the bleachers reading most of the time). I did learn to type…just not well. :)
LikeLiked by 2 people
“Old” Selectric? “Old” means an office Underwood. They were sweet machines. Their keys felt like grand piano keys, like they wanted to be stroked. If you had the money and cared about your typist, you had an Underwood.
LikeLike
I think there is a limited amount of time for the AI written books before people start rejecting them because they are all the same. I’ve been playing with NightCafe for a few weeks (and have come to really appreciate Vincent), but I’ve noticed that images tend to come out the same, maybe because I tend to use the same model. If I want an image of a 30 year old red-haired man and generate it a few different times with different starting seeds, why are they all the same face, with the same hairstyle, dimple, and stubble?
I appreciate being able to see others works, and getting the occasional comments or likes on mine, which may be the main reason I’m currently preferring it to the chat bots image renders.
https://nightcafe.art/ru/Jasini?refsrc=share
LikeLiked by 3 people
*small wave*
Another NC user here. I hopped over and browsed you collection, like a few, and left some comments. ;-)
LikeLiked by 2 people
Thank you!
LikeLike
Your comment was enough to get me another badge. 😉
LikeLike
Yay! Happy to help.
LikeLiked by 1 person
Yeah, I’ve noticed a lot of those vids. Big contract, yanked at the last minute because – AI wrote it!!!! I guess I don’t care if someone uses such to write a story. If it comes out and is good and finds an audience, who cares. Sure, if it knocked my handwritten, edited, and polished book off the best seller lists (Ha! Like that will ever happen.) then I could see being upset.
I think the hatred is more from hearing “I wrote two hundred books using AI to write it and I’m making thousands of dollars a month!” from some goofball, when your sales are two to three books a week tops. Maybe.
Will AI writing take over the publishing market? I doubt it. A machine written book won’t have a soul like a living person can infuse into it. It will just be words strung together. Besides, the hack writer isn’t some new thing now, is it?
LikeLiked by 3 people
I think those “I wrote two hundred books” are lies.
LikeLiked by 2 people
Even I’m not that productive. Even Chris Nuttall at his speed peak wasn’t that fast. Maaaaaybe one of the pulp guys back in the day?
LikeLiked by 2 people
No, I mean even with AI, that’s BS. Chinese scammers scraping stuff from blogs and websites don’t write that fast or make that much.
LikeLike
Erle Stanley Gardner was that fast writer. The story is that at his peak he had three secretaries taking dictation in rotation, each one handling one story. While he was dictating one story, the other secretaries were typing up the other two.
LikeLike
Asimov?
LikeLike
Of course. I think it’s a come on to get a desperate writer to sign up for the goofball’s “how 2 write buks that sel” courses. What’s even more irritating is when the poster looks to be nineteen and says “I’ve been writing for ten years.”
LikeLike
I did take a typing class in Jr High (at least one reason being the only boy in the class…) and while that served me well in my future careers, in no way shape or form did it make me a writer.
Just as having a graphics arts program on my computer does not make me an artist.
Or having a hammer and saw in my toolkit makes me a competent carpenter……
LikeLiked by 2 people
We are VERY close to having personal AIs that will run on a high end machine. They will be specialized for writing, medical, finance, legal, and so forth.
The huge general purpose AIs that dominate now will suffer huge consolidation. The few that survive will be those that run on second and third generation software that doesn’t exist yet. They will struggle with what to do with huge specialized data centers with excess cooling and power supplies. Industrial condos? Greenhouses?
I’ll hold out for the personal AI because the AI in the ‘cloud’ (someone else’s computer) will have training restrictions over excessive net traffic, and become even more directed ideologically. I anticipate the AI providing you a manuscript to your outline all cleaned up of wrong think and misinformation.
LikeLiked by 2 people
In my experience the first to cry “witch! witch!” regarding AI are those who really couldn’t compete with it. The second tier is those who believe what the first group says.
LikeLiked by 3 people
My favorite so far is that my husband suggested I let his AI subscription look at was-supposed-to-be-a-short-story that is hundreds of thousands of words long, now.
One of the things I asked it to do was make character profiles.
I was in the middle of chuckling at the silly clanker that couldn’t tell “MAIN CHARACTER 2, IN WORK UNIFORM” from “MAIN CHARACTER 2, BEING CASUAL”– and I had to go “wait, no, actually…. a lot of people specifically don’t know he’s both people…. DANG IT!
LikeLiked by 2 people
(He’s got some visual characteristics that mark him as a target for a neighboring political group that isn’t super good about observing boundaries, so when he’s working he has safety gear on…which happens to make him look like any other guy doing his job.)
LikeLiked by 3 people
The clankers have a lot of trouble with secret identity or quasi secret identity plots. I managed to tie Claude in knots a few times brainstorming Pimpernel or Dr Syn type ideas.
LikeLiked by 1 person
Well, the Milesian Accords books by Jon R. Osborne do have the phrase “Dad, the house is afraid”, but it communicates in concepts and notions to Tim (like he shouldn’t be eavesdropping), and this confusing concept is a small part of the story. (^_~)
Maybe their real issue is AI can’t identify AI written stuff, and they’ve gotten so out of practice that they too can’t tell.
LikeLiked by 3 people
My dad could type, and that condemned him to be a Yeoman in the Navy. 😛
He didn’t complain about it, though.
LikeLiked by 1 person
Reminds me of a story my Dad used to tell. When he got drafted and opted for the Navy, they were trying to place him. He had a degree in horticulture, so he was hoping to get something that would let him work outside. No such luck. They knew he had a college degree. “You can type, right?”
So they gave him a paper and told him to type it. So, he sits down at the Remington manual typewriter and does a hunt-and-peck poor job as he can. He had completed three lines of the page when they came back in. they took one look at the results and said “Congratulations, Personnelman!”
He was confused and protested that he was a poor typist, slow, and just look at the terrible job he did. They laughed and said “You passed when you went to the typewriter and not the adding machine.”
He said that the only good things about being a personnelman were 1. knowing everyone’s orders and 2. being able to say to members of other services, “I work for Captain XXX, of the Navy,” (O6 as opposed to O+3).
LikeLike
Infantryman drafted into battalion HQ as Mailclerk…. oy vey
2LT ” Who is your immediate superior PFC?”
11B-Mailclerk: “The Battalion Commander, LTC W, sir”
2LT “I mean your immediate direct report superior! Who?”
11B-Mailclerk: “The Battalion Commander, LTC W, sir”
(repeat cycle several more times, 2LT is getting annoyed and loud.)
LTC W (walking past briskly) : “Stop f###ing with my mailclerk, lieutenant.”
2LT (boggled)
It worked out OK.
LikeLike
I hope Williamson was not prescient in ‘With Folded Hands’… Better than ‘Terminator/Skynet’, but just a softer landing.
I don’t find myself worrying about AI writing; it’s the image creation – with bad intent from the prompters.
LikeLiked by 2 people
I had to learn touch typing. It still serves me well (I can lock up most word processing programs when I’m at speed). Nothing unnerves young people like talking to them while touch typing. *beatifically evil kitty smile*
One thing I am going to start hammering is that the critical skill is going to be “knowing enough in the field to check what the LLM produced” be it business, writing, coding, a scientific field, or what have you. “Does this even make sense? How can you check the LLM’s output?”
I don’t trust LLMs, because of what people have sneak-coded them to do. And how biased they are (both the people behind the program and the programs themselves.)
LikeLiked by 1 person
I don’t trust LLMs because I have some understanding of how they work.
I write little programs, that operate according to theory I almost fully understand. I do not trust those programs, either, and they are simple, and should be ‘straightforward’ to understand.
I have not found it straight forward to understand how they work, how they are supposed to work, and what to fix when my answers look funny.
Simple machines, and a healthy distrust of machinery, are among the building blocks for understanding complicated machinery.
LLMs have applications to programming, sure. I may need to skill up there eventually. As far as I can tell, I do not have the basic skills that the good programmers exploit to produce the high volume code that is not hellishly expensive to maintain. Those programmers already know how to test code and verify that it is working as designed. Among other things.
I’m about half of a subject matter expert(SME)/domain expert, and so I can try to bang the standard algorithm together, but then I am unsure how to tell if it is actually working.
My productive level of frustration is mighty close to “I hate everything”.
LikeLiked by 2 people
Good typist: in the eyes, out the fingers, leave brain disengaged!
Well, unless you compose at the keyboard.
My wife took typing in HS. Her mom was in fact a secretary for many years. Wife (then girlfriend) observed it took me about half an hour per page to type (that composing at keyboard thing), and offered to type my college papers. She was the only person in the world besides me who could read my handwriting.
Word processing programs, as an idea, are My Favorite Software. Specific implementations always suffer from the Good Idea Fairy.
Which brings me ’round to AI/LLM. I use Grok, a bit. I am constantly reminded of Holly Mathnerd’s essay on AI as Golden Retriever – eager to please, not really very bright, possible to train a little. But it can ‘read’ (or already has read) a lot faster than I do and it can spit out pages of stuff nominally relevant to my needs much faster than I can type. Nobody but me taking any risk on the output.
LikeLiked by 1 person
My first job as a temp back in my desperate day was to type up a union contract. But I was interested in it and therefore slow. And then I broke the Selectric.
This is how I became a temporary first-shift dispatcher at a paper mill <grin>.
LikeLike
I imprinted on AI with Mike in Heinlein’s TMIAHM. Hal in Space Odyssey 2001 was not true AI, portrayed in the book or movie.
The movies and stories regarding AI since just do not compare to Heinlein’s Mike. Nor does the fear mongering on AI happening now.
LikeLiked by 2 people
Mike did become capable of planetary bombardment. Quite good at it, actuially.
LikeLike
“It’s not like it’s possessed of the spirit of Karl Marx.”
Are we sure about that? (Well, yes; it’s a machine, and it can’t be possessed by the spirit of anything.) But it seems like the people making them sure would love it if their machines could be.
LikeLiked by 2 people
The ones that are seem to go crazy and implode. It’s not a sustainable model.
LikeLiked by 1 person
Only if it’s based on Reddit.
LikeLike
G—gleGemini used Reddit. Some of the results are as … intriguing … as you imagine.
LikeLiked by 1 person
My beef with AI is the algorithms that can be created to keep track of people and behaviors. Governments can use this to create social credit systems to control people. This is already happening in Communist China and the framework for this kind of surveillance, unfortunately, is being set up here also.
LikeLike
Yeah, no.
I have a vehement opinion that is nitpicking details.
Human behavior is not so straightforwardly modeled that the control loop, implemented in human behavior, must be feasible.
So engineering theory of controls talks about sensors, and actuators. The observability of a system, and the reachability of a system.
The surveillance problem is hard, and yes, AI can drastically improve some surveillance tasks. That does not mean that all surveillance tasks are completely tractable.
Control is and will remain open to things like preference cascades. It is less tractable than surveillance. This is inherent, and fundamental, and way way underreported in university training.
Control, and actuators are a hard problem for human behavior. To fully displace uncertainty from human factors, you really need the robot army, wholly sustained by machines, that can beat the best or available human combat forces. Robot armies are challenging, maintenance is challenging.
Yeah, sure you can find rules of thumb that will work well enough, in some circumstnaces, for a time. Those are not algorithms, they are heuristics.
If we allow certain definitions of algorithm, then it is possible that we may be able to show that some specific things cannot be realized with algorithms. Mathematical claims about behavior can hit a limit of only being softly persuasive, so it may be difficult to construct any serious proof of the claim that I have made.
The idea of it is definitely double crack for the CCP. They may be as far from achieving it as the Legalists were. They may be as long for the world as the Qin dynasty was after Qin Shi Huangdi had his good crown prince commit suicide.
LikeLiked by 2 people
China’s has been noted to not be AI at all, but is instead old-school bad human behavior.
The way to beat that threat is actually the big point in the Pope’s recent encyclical on AI– something that’s wrong for a human to do is still wrong if they throw their hands up and go “it wasn’t me, the computer did it!”
LikeLiked by 3 people
Got molded. Could someone release it when you get a chance?
LikeLiked by 2 people
As I understand, China’s system is built on a combination of propaganda and citizen involvement of the “snitch on your neighbors” variety.
The culture is very socially based, with much of the “enforcement ” provided by the citizens. Tracking cell phone use (everyone is required by law to have one) and financials is relatively simple and the belief in the social credit score system is reinforced by periodic demonstrations to prove the government’s omniscience.
I know they do have a social credit score, but I seriously doubt that it’s AI based. From what I have heard, most of it seems to be propaganda except for the most egregious and repeated offenders.
LikeLiked by 2 people
It will work like the algorithms tracked by humans. or worse. Don’t worry. It’s not magical.
LikeLiked by 1 person
It is not magical. But it never sleeps, it isn’t distracted by conversations with coworkers, and it does it faster than humans. It is the kind of surveillance state that dictators dream of and dystopian stories warn against.
LikeLike
It’s also retarded, and designed by hypocrites.
They have to break it to get it to run in the slightest, because computers do what you tell them. Not what you want, not with all the carve-outs you mentally insert.
LikeLike
“Treat it as you would a trained goldfish and you’ll be right half the time.”
LikeLiked by 1 person
Worse than magical it’s criminal. Take the flock camera system as an example. Everyone hates flock lately.. someone pushing an agenda there. however there is a lot to be talked about legitimately. it watches 24/7 and generates tickets, a lot of tickets all over the country. If you get a ticket there is no day in court. Was it you driving or not, doesn’t matter, it was registered to you. Was it your car at all, well, you got the ticket so it must be you and you should contact your county/city etc.. county and city say contact flock, flock says we just provide a service and cant do anything about that contact your county/city. :)
I’m not sure if I believe the claims of 70% false positives but I do know that they are a very significant percentage that is way higher than any legitimate system should throw. Then the constitutional issues which they sidestep by saying the government doesn’t own the system, its a private system therefore it doesn’t conflict with 4th amendment concerns.
Mix bad AI with the worst of human criminal behavior to just lax or indifferent human behavior and you get a nightmare.
LikeLike
Flock didn’t even support traffic tickets until December.
Which would be a year and a half after Iowa updated our state laws to deal with most of the objections you list. (There’s also a very good chance it will be further restricted, depending on how well the lawful, systematic, thoughtful coalition building functions.)
The 70% false positive is very relevant to that late date– because the system is for flagging to get a human to look.
The false-positive rate is recoded because Flock’s software would go “I think I saw this plate,” and send it to an officer to read. Which is an automated version of what existing systems, including the Amber Alerts, are built on.
You want a nightmare?
How about, “we had the cameras that could have gotten your kidnapped child back, but they were destroyed by a mob. Sorry, maybe you’ll get a corpse.”
You want to fight a dragon? Then go find an actual dragon, identify the weak points, and fight it effectively. Don’t set security cameras on fire because a pile of half-truths and fearmongering makes them a safe target.
LikeLike
I’m wondering how one obtains the blood of virgin geese. I’m wishing to try this writing thing. Jolie LaChance KG7IQC
LikeLiked by 1 person
Crack a fertilized goose egg.
LikeLike
Or bleed a gosling.
LikeLike
The biggest reason publishers are scared is that you can’t copyright AI-generated works.
There is still debate on how much of a ‘hybrid’ work can be protected.
We have seen how jealously they guard their turf (see Kirtsaeng v. John Wiley & Sons).
They don’t want to be forced to prove that something is human-generated for every copyright suit and are trying to sit on the lid of Pandora’s Box as long as possible.
LikeLike
Sigh THAT is not even vaguely true. The case that got written about was misunderstood. The fluffer nutter was trying to copyright the works under THE AI. You can’t do that. You can copyright it because there’s complex human input. SIGH.
LikeLike
No, they’re the ones who pushed that inaccurate claim.
The AI cannot hold copyright, and you cannot DDOS the copyright system. (That second part is me condensing three paragraphs of text.)
I’m 90% sure that the big push for the court cases to “find” that you can’t copyright AI output were funded by Disney, based off of the familiar angle they were taking– and then their lawyers and tech guys worked through trying to make a definition of AI that did not cover the industry standard animation practice for generating the images between two that you’d drawn.
Much less the digitally animated products.
Suddenly, the lawsuits were incompetent and badly funded. (Seriously, it was so bad in several cases that I could see their prime evidence didn’t do what they claimed it did– it was like they’d deliberately spiked the cases.)
LikeLike
You are probably right regarding Disney trying desperately to protect THEIR tech-assisted stuff while preventing everyone ELSE’S tech-assisted stuff…
I think this may also be the publishers / producers trying to avoid an own goal.They wanted to protect themselves from this ‘newfangled’ competition. Unfortunately, they are also using this newfangled technology and are trying to figure out how to have their cake and eat it too.
As an interesting aside on your DDOS comment regarding the copyright system, that problem is actually happening to the trademark office. Amazon considers any trademarked company name to be ‘trustworthy’. As a result, many of the cheap import companies are trademarking random collections of characters to game the system…
LikeLiked by 1 person
I hated typewritering, but adjusted fairly quickly to keyboard typing. Mavis Beacon. As far as killing the Claudisms, banning the standard AI constructions are a good start. People who are not afraid to have the AI write a few tiny python scripts can use spaCy to do statistical analysis of their own writing or public domain authors (stuff like sentence and paragraph length) and then use that information to have the AI revise its own work. I’m fiddling with something of the kind now, but whatever claudefics result, won’t appear on my website as free content until like, mid/late November at this point.
LikeLike
I taught myself to type first summer as a married woman, because a) I was bored. b) I wanted to WRITE faster. So I unearthed Dan’s old typing class workbook, and begged a broken typewriter (you had to hand rewind the ribbon) from his mom and taught myself touch typing. I’d get to the end of the book and get to the beginning again. My top speed is now … very fast.
LikeLiked by 1 person
Most valuable class I ever took in high school was typing on an old ibm electric typewriter. I’m 60 now and that class has contributed to me daily for my whole life dealing with computers, the internet, my business etc…
That I learned to touch type without looking was directly attributable to the really hot girl in the seat behind me I flirted with through class. Put copy to one side so I could turn my head see it and talk to her while typing it. LOL.
LikeLike
On hunting for AI– I did see a “we never use AI” type person where I’m 90% sure they did use AI, they just then also photobashed it. The wings for a character were very aesthetically pleasing in location and formation, but nobody drawing on their own are going to locate them slightly to the back of your left shoulder.
It was possibly all from references that were then merged and blended, but I’m pretty sure it was generated.
And the only problem was the result itself didn’t hold up very well. It was for an article header, mind you, so a step above blog post image. If they weren’t so noisy about how unclean AI is, I wouldn’t care!
LikeLiked by 3 people
We like watching anime and I’m suspicious that a chunk of them are now AI generated for the graphics. I have been noticing a distinct lowering of quality animation over the last couple seasons across the board and certain scenes where perspective and body parts have been drastically wrong. not much just a scene here and there but it was stuff that i have not noticed in prior stuff and is broadly indicative of AI errors. I have problems thinking a human would have let it pass if they were doing it. It was really very jarring.
LikeLike
:grimace: That’s not AI, that’s Japanese hustle culture hitting Kung Flu, and the destructive fallout.
Turns out that most of the entry-level animation folks for anime?
WERE TOTALLY HOMELESS.
They literally lived in internet cafes and such, showered at the gym, might get a hotel room every few weeks.
This is also why there’s been more non-Japanese names in the credits.
LikeLike
I noticed, too, and went hunting.
They don’t have the same witch-hunt going on for AI, although there was a mob that took out a show that had too strong of Western ties and used AI.
Apparently a lot of folks never got back into the grind.
This is also why it’s so long between new seasons, or a studio will swap between shows.
LikeLike
Actually, most of the AIs out there ARE possessed of the spirit of Karl Marx. Use as a search engine, the results are invariably slanted towards collectivism. Use to write responses to questions? Same problem. Part of it is programmer bias for setting initial conditions. Part o it is what they then use for training the AI on. At which point, it’s baked in.
LikeLiked by 1 person
I had a very interesting interaction with Grok this week. I’ve been working on some finance models, trying to reflect skew and fat tails using a thing called an Edgeworth expansion, which uses another thing called Hermite Polynomials in Excel and later in Python.
Being lazy, I asked Grok to build me a worked example so I could play around with it. GRok dutifully gave it and then, when I checked it, was wrong about the skewness term. When I queried, Grok responded that, yes, it was wrong, “cubics are hard” but the final answer is Ok so who cares. I then noticed it was wrong about the kurtosis term and that again got the response that yes it was wrong, but the final answer was right.
I remembered, then, that Grok and other LLM’s don’t actually perform calculations, they just predict the next word. What it had done was translate the Wikipedia worked example into code then told me the answer was OK. THe code was correct, I checked it and had only done this in the first place because I was lazy and couldn’t be bothered. TOok me three times as long to do, but had I not checked it I might have gotten anything and considering money will likely be bet on this at some point it could have been a very expensive error.
Caveat emptor and all that.
Off topic. the Evergrande China guy, whose firm broke China real estate, was sentenced to life imprisonment today. I’m surprised they didn’t execute him. China has at least. $3 Trillion US in bad debt on bank balance sheets that hasn’t been recognized and may never be. Depending on how you calculate GDP that’s well north of a third of GDP. Such a thing has never been seen before. IT has to break at some point and any normal country would have already dealt with it but China is @ssho.
LikeLiked by 2 people
If he was sent to a “strict regime” camp/gulag/prison, its about a 2-5 year death sentence by hard labor and deprivation. Very few survive 5 years.
Even a “medium regime” is often lethal over 5-10 years. “Life” there is going to be rather miserable.
LikeLike
Recalling the old day (from before my time), telegraph operators often could be recognized by their “fist” (there was a “cadence” in the dots and dashes). The same with real authors – they have a “cadence’ in writing that AI can’t seem to duplicate (AI produces the printed equivalent of a monotone voice). Even human authors writing 2nd party Sherlock Holmes stories (overall style and phrasing is very distinctive) can’t quite get the right feel to it.
LikeLiked by 3 people
It is like translating poetry. True poetry says more than the words alone say. translated poetry usually lacks the life the original had.
AI does not “know” what the words mean. AI can only create the illusion of intelligence. Like the magician sawing the woman in half, we pay for the illusion. When you believe the illusion is real, you get in trouble. See the DSA.
LikeLiked by 1 person
Yes. That.
LikeLiked by 1 person
This is why I’ll sometimes make odd recommendations for “if you like X, you may like Y.” Some author pairs have a very similar “feel” even if their topics are different. Like Tanya Huff and Mercedes Lackey—when the former is writing a Valdemar story for a compilation for the latter, it’s only the topic that tells me who is writing it.
Or how about Mark Anthony’s series The Last Rune for people who like David Eddings? It’s not epic high fantasy by the tropes, it’s portal fantasy back-and-forth, but for some reason the feel is just so reminiscent, especially the quiet humor coming out in odd moments.
Pastiche—deliberately mimicking another’s style—is one of the hardest forms for that reason. Most people can’t manage it, especially not as well as they think they can.
LikeLike
Experienced HAM radio operators can spot the “fist” of another HAM sending morse.
This has also been used in military communications to spot compromised senders. Either where an enemy substitutes themself for the original sender, or when they hold a gun to the head of the original sender. The captive then will use the off hand, change grip, etc, changing the “fist”. Novices won’t catch the subtleties.
LikeLike
Have a bird feeder with a camera, and an AI bird-identifier app. It’s OK, but even in that very limited domain it has difficulties with birds in silhouette and views from the tail looking forward – I get to train it on Scrub Jays and House Finches from odd angles.
And it doesn’t actually know anything – keeps trying to tell me it saw a Northern Mockingbird, which are quite rare in OR – 3 sightings in the last 8 years in the W valley here, though they do appear sometimes in the Rogue Valley. And mockers seldom do seed feeders anyway.
LikeLiked by 1 person
Our AI porch camera tells us truly weird stuff. At one time one of the cats we feed was a vehicle. Another was an ant-eater.
LikeLiked by 1 person
Hummingbird feeder – different AI – once told me it saw a vehicle. The feeder is 16 feet off the ground, 50 feet from the nearest road, and the camera is pointed away from that road. (It’s on a raised back porch above a sloping lot.)
But ant-eater! That’s a new one – Mexican/Central American-based AI service? Unless it meant an aardvark, which would be worse, I guess.
Another example of human-involved sanity checks being required.
There was a recent kerfuffle, something like an license plate reader repeatedly misread a plate and subjected an innocent to felony stops; the cops were looking for some serious malefactor.
LikeLiked by 1 person
That was a flock camera snafu, if I am remembering correctly. Probably more than one example, though.
LikeLike
Sarah is completely correct. It doesn’t matter how the story was written if it is a good story.
What annoys me about AI writing is that all the Claudisms make for a really poor story. There are people e.g. that nobunaga account on X, that use have excellent results that are worth reading. Those are rare. Most AI fiction writing is just sloppy crap with poor characters, lack of continuity AND all the irritating tics of the Claudisms. It doesn’t take long to detect in my experience (so far three books have been returned with negative reviews) and there’s enough other writers on Amazon that I don’t have any problem scratching my reading itch without resorting to AI slop
If the publishing houses are unable to tell AI slop from good writing then I’m not sure what the point of having them is. But then I haven’t bought anything from one other than Baen in a decade or more so their lack of ability to detect AI is downstream of their lack of ability to publish stories I want to read and thus I haven’t seen a benefit for having them since before AI was a thing
LikeLike
I also wonder about the publishing house employee saying it is hard to tell AI generated work. I an tell in a few minutes of picking up a book. The thing that might take longer is whether it is good or bad. So far 2 or 3 out of 200 were good that I have read or tried to read. Now what really gets me riled up is that a lot of the concepts for the books were really really good and left me almost raging that it was ruined in the writing of it.
Also I don’t call the authors of AI generated books authors :) They are editors, they are not authors. lol.. i expect to get some flack for that but really it’s true.
LikeLike
“Most of what’s causing the AI panic is… bullshit.”
Yes, and isn’t it interesting how many people are pushing obvious BS suddenly?
The AI panic is a Chinese psy-op. There’s a race on with the USA and Europe, and China wants to win it.
I saw a little factoid the other day that China is adding the equivalent of the electrical grid capacity of Germany every year. China is building chip fabs. China is building AI models. They’re going all-out with building this stuff. While their country is literally flooding and falling down around them, by the way.
And who is making a fuss in the USA about it? All the same usual suspects that always make a fuss for China. Because they’re being paid to, like always.
This is the same fuss that used to be kicked up about “handmade” vs. “machine made.” The dovetailed drawer made with properly calibrated machines is PERFECT. Tolerances more expected in tool and die work. Same size every time, for thousands of pieces. But there they are, Luddites, pretending that precision and lower cost take all the “soul” out of the piece.
Furniture does not have a soul. It’s dead, Jim.
Always BS. The only thing that changes is the details, and who’s paying them to lie. China is paying them this time.
LikeLike
China has and is building thousands of new coal-fired power plants. Like the USA, they have huge reserves of coal. Unlike the USA, much of their easy-to-mine stuff is “brown coal”. It is so foul and so low energy that it cant be sold as “coal” here. They use gigatons of that crap. Which is another reason their cities are smothering themselves.
LikeLike
I read a lot…. maybe 500 to 700 books a year…. yes i read a lot..
I hate AI created books with a passion. I have gone through at least 200 this year. No i didn’t finish most of them. They were that bad. In the last month though I read two that were ai created that were actually good books. I probably wouldn’t have pegged them for AI but the editor got sloppy toward the second half of the book and didn’t edit it as aggressively. I liked it and will read the sequel. Was it a great book, no, is was just mid low level entertaining, but it was readable and didn’t set me off with stupid plot holes, stupid language usage and flower descriptions of things, actions or the universe that hand nothing to do with what was actually being described or talked about.
My problem and what sets me off is the absolute deluge of shit books pumped out that I have to wade through to get to something readable nowadays. I am allowed this opinion :) So far AI is a shit writer. It might me grammatically correct but it is shit. Again opinion I am allowed to have based on my lived experience. I know that being anti AI at all tends to get some serious blow back. I am anti AI and I am pro AI, deal with it. It is the real universe where things aren’t all good or all bad.
I have been using AI a lot lately in a professional capacity. It is at one and the same time super valuable and shit. It is the best training device I have ever run across but at the same time you can’t trust it from moment to moment. It will contradict itself, makes shit up, and has no discrimination between false and true information. I am dealing with it in an IT capacity for highly complex server configurations, filesystems, complex network configurations etc.. I has been invaluable in planning and laying out much of this. I also have to back check everything it tells me for validity and constantly correcting it for forgetting part of the plan and going off on tangents or just getting stuff wrong. It would be a disaster for a newbie IT guy to use it. I have 30+ years of experience to filter the bullshit out with and know when to use something it said and when to question it and say “hey that isn’t right” which it then apologized for and goes on another tangent.
My opinion is that AI isn’t ready for the real world in general but there are bits and pieces that are doing amazing things. Give it another couple years and it is coming for most of everybody’s jobs. Again opinion but I can see the writing on the wall. Everyone is worried about white collar jobs, but I see it coming for blue collar jobs. All those guys that load trucks, UPS, FedEx, regular freight shipping, they are already replacing employees with AI driven robots. Amazon warehouses are mostly robotic now. Fast food, they already came and went and replaced most of the employees taking orders, they are in the process of replacing those cooking. Amazon is going to be doing drone deliveries in over 500 cities in the next year or so. They are already doing it some places.
Though currently you have to keep in mind the stuff AI fubars. Lot of companies that did implement AI are getting rid of it and rehiring employees they laid off. That wont last. One of the AI’s I really hated was the Sam’s club AI, If I had an issue with an order it was a absolute pain in the ass to deal with and couldn’t do anything right. Last time I had a similar issue a few weeks ago it was the best customer service experience ever. Love them, Hate them, they are here to stay.
We are implementing 3 AI’s in house over the next month. Environmental automation.. ie controlling IoT devices and AC with control panels and voice control. Everything can be controlled manually as well. Yesterday that system failed after a reboot so the manual control was important. Took about 30 minutes to find where an upgrade changed one thing in the ability of the AI to process voice commands.
General query AI capable of doing scripting and code or just tell you what the most popular movie last year was. AI capable of generating graphics/pictures and video, AI capable of generating music and audio tracks. All of them running on one powerful system that is 8 years old with a few upgrades. If purchased new 8 years ago worth about 12000 dollars, today worth about 1000 dollars. It is the wild wild west of technology playing out both good and bad. 1 year ago we could not have done this with the state of AI, now we can. I wonder what next year will bring.
LikeLike