AI written books

[Quoth]

Originally Posted by DiapDealer

Predicting what the worst humans might do before they do it?

Ha!

Do we really need a computer for that? It would also be a failure*.

I've not watched Minority Report, but you can't really know even what some well known and documented people might do in advance and it's not reasonable to arrest and jail people for what they might do, though it's been done. Also facial recognition systems have been misused to arrest people or ban them from shops or venues without any human checking or advance challenge. Given bias in training data and poor performance against the majority on humans (who are not pale-skinned) a human verification and investigation should occur first.

* Like Asimov's fictional (at the time) Psycho-history it's nearly impossible to be sure of an individual's future actions, but in a general population trends can be forecast more accurately. But we don't need AI for that.

[DiapDealer]

Sorry. I know that you feel you have your finger on the pulse of what the human race might, or might not, need from all things AI (I also realize most of your last post was intended humorously) -- and I don't disagree with all you have to say on the subject. But I'm not quite willing to accept your opinions on humanity's "needs" (or fears) for AI as gospel just yet. Humans don't always know what they need until after they need it. You, me, and the guy next door, just don't know what we need AI for yet. Simple as that. Ima wait and see. My number just came up today to take Google's Bard for a beta spin.

[Quoth]

Originally Posted by DiapDealer

Sorry. I know that you feel you have your finger on the pulse of what the human race might, or might not, need from all things AI

No, I don't feel that at all. Certainly some aspects of AI over the last 50 years approx are useful. OCR was once thought to be AI. Maybe it still is. It's certainly much better now than the groundbreaking launch by Ray Kurtzwiel in 1974 (nearly 50 years ago!). I think he works on AI in Google/Alphabet now. OCR'ed stuff still needs proofed by humans.

I'm just suspicious of hype and of course as Charlie Stross might say on the subject "follow the money".

If all this AI is so brilliant how come it's as difficult to spell check and grammar check as the mid 1980s? Proofing is aided by being able to read and annotate on eink.

I'm not fearing any kind of Skynet or Minority report scenario, just offering an antidote to hype.

Certainly the current crop of chatbots are more entertaining than the ones in the 1960s to 1990s.

[Hitch]

Originally Posted by Quoth

No, I don't feel that at all. Certainly some aspects of AI over the last 50 years approx are useful. OCR was once thought to be AI. Maybe it still is. It's certainly much better now than the groundbreaking launch by Ray Kurtzwiel in 1974 (nearly 50 years ago!). I think he works on AI in Google/Alphabet now. OCR'ed stuff still needs proofed by humans.

I'm just suspicious of hype and of course as Charlie Stross might say on the subject "follow the money".

If all this AI is so brilliant how come it's as difficult to spell check and grammar check as the mid 1980s? Proofing is aided by being able to read and annotate on eink.

I'm not fearing any kind of Skynet or Minority report scenario, just offering an antidote to hype.

Certainly the current crop of chatbots are more entertaining than the ones in the 1960s to 1990s.

Waaah, the first of what I'm sure will be many AI-written and created books has just been thrown over the transom; I'm waiting on a reply to our quote now.

The author wrote a humorous 1-2-3... sort of Guide for adults--by which I mean it's illustrated but it's not a kids' book. Anyway, all the illustrations, every single one and they account for half the content--are AI-generated. They're GORGEOUS and charming.

This doesn't portend well for artists.

Hitch

[DiapDealer]

Originally Posted by Quoth

I'm not fearing any kind of Skynet or Minority report scenario, just offering an antidote to hype.

I never thought you were worried about Skynet. But with your last several "we don't need AI for that" proclamations, it's certainly not out of line to assume you might think you're making objective statements RE mankind and AI. There's "offering an antidote to hype" and then there's dismissing all other opinions on the subject as incorrect and/or misinformed. *shrug*

[andyh2000]

Originally Posted by Hitch

Waaah, the first of what I'm sure will be many AI-written and created books has just been thrown over the transom; I'm waiting on a reply to our quote now.

The author wrote a humorous 1-2-3... sort of Guide for adults--by which I mean it's illustrated but it's not a kids' book. Anyway, all the illustrations, every single one and they account for half the content--are AI-generated. They're GORGEOUS and charming.

This doesn't portend well for artists.

Hitch

Do the illustrations have a consistent style?

Andrew

[Hitch]

Originally Posted by andyh2000

Do the illustrations have a consistent style?

Andrew

YES, they do. I wish I were a bit unscrupulous; I'd post a few samples, but I can't do that to a prospective dlient or anybody else, unfortunately.

That's probably what freaked me out the most, lol. But on many of these AI-driven illustration/composition sites, they do ask you for the "style" you want something done in and you can get very, very specific. I have no doubt tht he spent...I can't imagine, how many hours on the first one, making it just so, getting it JUST right. It's a remarkable piece of work, if you consider the march of technology, right?

Hitch

[andyh2000]

Originally Posted by Hitch

YES, they do. I wish I were a bit unscrupulous; I'd post a few samples, but I can't do that to a prospective dlient or anybody else, unfortunately.

That's probably what freaked me out the most, lol. But on many of these AI-driven illustration/composition sites, they do ask you for the "style" you want something done in and you can get very, very specific. I have no doubt tht he spent...I can't imagine, how many hours on the first one, making it just so, getting it JUST right. It's a remarkable piece of work, if you consider the march of technology, right?

Hitch

I'm torn about image generation AI models. On the one hand I can see why artists, photographers etc are outraged that their work has been used without consultation or recompense to train things which then make money for other people and/or potentially replace the original creators. On the other hand my very limited knowledge in the area wonders how it's different from every jobbing artist out there who studies lots of other people's work, learns from it and maybe starts out imitating styles before developing their own. I guess I maybe disapprove of people who ask the AI to produce a picture "in the style of <artist>" and I'm kind of neutral to approving if they describe features they want in the output image without mentioning people. For commercial or public use that is - what people do with it for their own personal amusement is a different matter.

Andrew

[Hitch]

Originally Posted by andyh2000

I'm torn about image generation AI models. On the one hand I can see why artists, photographers etc are outraged that their work has been used without consultation or recompense to train things which then make money for other people and/or potentially replace the original creators. On the other hand my very limited knowledge in the area wonders how it's different from every jobbing artist out there who studies lots of other people's work, learns from it and maybe starts out imitating styles before developing their own. I guess I maybe disapprove of people who ask the AI to produce a picture "in the style of <artist>" and I'm kind of neutral to approving if they describe features they want in the output image without mentioning people. For commercial or public use that is - what people do with it for their own personal amusement is a different matter.

Andrew

Sure, but what's the difference, really if they go to some working student or even graduated, degreed artist and ask him or her to make a work "in the style of..."? That happens in working commercial shops all the time. I mean, ALL the time. (e.g., the Dali clock. How many times have we seen that "homages" to death?)

So...it's a conundrum, it is. I feel for anyone whose job is lost to technology, I truly do--I feel that pinch myself. Things that you couldn't do without some markdown (at least) knowledge 10 years ago, in an ePUB, etc. software now does for you, however imperfectly. We have software that does have a learning curve, but has such a substantial HELP system built in (like Adobe's InDesign) that offshore would-be book designers can pick up a copy of InDesign on Monday, and hang out a shingle on the wall of their Fiverr shop, by Wednesday, saying "Deziner is in and available for book Dezine help." And how does the publisher know that they're getting a layer of top-flight software layered over 3 cake layers of no-training, etc., when all they see is that top layer?

Anyway, not to digress, but is it really different to ask a human to do it, than asking a machine? Or is that just an excuse we're all telling ourselves? (Genuinely asking, not...not debating argumentatively!)

Hitch

[andyh2000]

Originally Posted by Hitch

...
Anyway, not to digress, but is it really different to ask a human to do it, than asking a machine? Or is that just an excuse we're all telling ourselves? (Genuinely asking, not...not debating argumentatively!)

Hitch

I think to me it's the scale and speed that feel different rather than some ethical distinction. The ML model is trained on every single image its keepers can scrape from the internet and can output say 10 images per minute on a decent home PC covering everything it's seen. An aspiring human forger spends a considerable length of time studying just one artist and I'm sure even the best digital painter takes longer than 5 seconds per picture. So it's a question of degree rather than kind.

People who think there's something other than large scale long term reinforcement learning of connections going on in our brains might disagree. Every advancement I see in AI just pushes me further towards the opinion that we're the latest in a long evolution of meat machines, nothing special.

Andrew

[Hitch]

Originally Posted by andyh2000

I think to me it's the scale and speed that feel different rather than some ethical distinction. The ML model is trained on every single image its keepers can scrape from the internet and can output say 10 images per minute on a decent home PC covering everything it's seen. An aspiring human forger spends a considerable length of time studying just one artist and I'm sure even the best digital painter takes longer than 5 seconds per picture. So it's a question of degree rather than kind.

People who think there's something other than large scale long term reinforcement learning of connections going on in our brains might disagree. Every advancement I see in AI just pushes me further towards the opinion that we're the latest in a long evolution of meat machines, nothing special.

Andrew

Yes, but see, that's peripheral, then, isn't it? I mean, to test any question, we have to take it to its ultimate limits, yes? So, if, for you, it's the speed and all that, when would they approach parity?

If the machine took 1hr, 55 minutes and the human took two hours--what would your thoughts be then?

Hitch

[andyh2000]

Originally Posted by Hitch

Yes, but see, that's peripheral, then, isn't it? I mean, to test any question, we have to take it to its ultimate limits, yes? So, if, for you, it's the speed and all that, when would they approach parity?

If the machine took 1hr, 55 minutes and the human took two hours--what would your thoughts be then?

Hitch

I think I'm agreeing with you that there isn't much ethical difference in getting a machine to copy someone else's work/style versus paying a person to do it. If they both took the same length of time then you'd still use the machine as I'd expect it to be cheaper (no hungry kids to feed).

Andrew

[Hitch]

Originally Posted by andyh2000

I think I'm agreeing with you that there isn't much ethical difference in getting a machine to copy someone else's work/style versus paying a person to do it. If they both took the same length of time then you'd still use the machine as I'd expect it to be cheaper (no hungry kids to feed).

Andrew

Yes, which then casts the entire discussion into stark relief: does it matter, other than preserving an artist's livelihood and is this societally required, or is it sort of "tough luck," just like when robot automation changed factory life and all that? And everybody has to simply 'adjust'? Does not seem to have done Detroit much good.

Hard to know.

Hitch

[gmw]

Originally Posted by Hitch

Yes, which then casts the entire discussion into stark relief: does it matter, other than preserving an artist's livelihood and is this societally required, or is it sort of "tough luck," just like when robot automation changed factory life and all that? And everybody has to simply 'adjust'? Does not seem to have done Detroit much good.

Hard to know.

Hitch

Well, realistically people will have to adjust. I'm hard pressed to find examples where the march of "progress" has been stopped just because a few livelihoods are being trampled on. I think most recognise that and it's part of the fear, "Am I going to be on the hit list?"

Publishers are obviously concerned at the moment of being "caught out" accidentally publishing AI produced material (without acknowledging it as such). But as your posts have already mentioned, much of the time it will not be as clear cut as that. It will be some portion, or some level influence or assistance.

Writers rarely acknowledge the software or hardware they use (although I do seem to remember reading someone's ode to their pen). However, AI threatens to change the ratio of human vs machine contribution. Our history with copyright and its related "fair use" doctrine, will inevitably excite questions in at least the writer's and publisher's minds of who is the actual creator. Did the machine create the work and the human merely adapt it? Or did the human direct the creation and so must be considered the creator?

We might look to photography. There is generally no question that some photographers are creators of art, even though the end result is arrived at entirely through technological means. And the same laws that protect such art also protect the family snap taken with little thought, and only the effort required to lift their phone and touch the screen, even though it might be argued that much of the artistic value (lighting, contrast etc.) has been chosen entirely by the machine. And the situation has been muddied further by animals taking their own selfies.

So in some respects the problems presented by AI generated text are not new. What changes, perhaps, is the idea that the deliberately creative part - up to now assumed to be the minimum human contribution - may get usurped by AI. It's not going to do that yet, but its attempts are close enough to make us fear that we humans will caught out as not quite so inspired and creative as we thought.

[j.p.s]

Originally Posted by gmw

I'm hard pressed to find examples where the march of "progress" has been stopped just because a few livelihoods are being trampled on.

Georestrictions on ebook sales.