Well, it looks creepy and kind of interesting, but . . . I don't see a Beholder.

View this comment - View article - View full comments

Quoting: CyrilReading your answer, I'm wondering if we speak about the same thing.

I'm just making the point that not everyone wants to get into this stuff, and that that is not a moral failing as computer people so often seem to imagine when they, say, stick out their tongues at people who do not share their tinkering interests.

And for practical purposes, I generally do not find myself needing to do this. If I do find myself needing to do this, for software I actually have a significant need for, yes, I consider that a failure. The ideal is "just works" not "have to faff around". Is that fair? Maybe, maybe not, but that's life in the big city; it's the expectation casual computer users are gonna have.

Just to be clear, I started using Linux back around 2000, and in those days I did have to learn how to do some command line stuff, most of which is no longer relevant (I used rpm-based then) or I have forgotten. But I was not happy about it; I used Linux despite having to muck around with it. If Linux had stayed until now the way it was back then, I would have given up on it.

View this comment - View article - View full comments

Quoting: Kimyrielle

Quoting: Purple Library GuyYeah, but the point is everything an AI produces is somebody's Yoda or Harry--maybe two or three somebodies mixed together if you're lucky.

This is widespread, but still false assumption. A good approximation to 100% of all characters generative AI will draw have never been drawn by anyone before.

That's a strong statement. How do you know?

View this comment - View article - View full comments

Quoting: Grogan

Quoting: Purple Library GuyYeah, but the point is everything an AI produces is somebody's Yoda or Harry--maybe two or three somebodies mixed together if you're lucky.

Everything everybody creates has elements of something somebody else has created. It's how humans progress, by building on the work of others.

Right now AI is rudimentary, but soon it will be doing much more than mixing and matching samples it has learned.

Your first point, yeah, true enough. One of the most "creative", out there, poems ever written, "Kubla Khan", a guy wrote a book dissecting everything Samuel Taylor Coleridge had been reading over the previous few years to find the sources of all the elements that went into the poem. But he understood what those sources were, and did intentional, thoughtful things (and unconscious, stewing in the background of his mind things) to them when he brought them into his poem. I think there's a difference between that and what these art programs do.

On your second point, I'm not sure it will. The stuff they're calling "AI" right now isn't the result of some conceptual breakthrough or anything. It's just autocomplete on a whole lot of steroids. If you look at the output of ChatGPT, there are "tells" every so often that it doesn't actually understand what it's saying--or rather, that there isn't really an "it" to do any understanding. It'll say things that any fool can tell make no sense together, things that contradict each other and stuff. And the scale of the datasets is so large now that I think they've pretty much reached the limits of the "get a bigger hammer" approach--there isn't much more data to get.

Don't get me wrong, the results are impressive. It's quite surprising what ChatGPT can pull off. But we assume it will continue to get a lot better because we think of it as "artificial intelligence" that just isn't very smart yet. If that were so, it would have lots of potential to improve, it would just be a matter of learning to understand better. Problem is, it isn't artificial intelligence, it's just software, and software tends to plateau in its capabilities at a certain point. How much better have word processors or spreadsheets gotten in the last dozen years? The current art and chat oriented "AI" programs are not the product of a brand new concept; the ideas behind them have been maturing for some time, and getting trained on larger and larger datasets for quite a few years. They may not have a whole lot better to get. To get very much better, they would have to actually understand what the words mean that they're putting out--and they don't, and that would be a completely different kind of breakthrough that nobody currently has a handle on.

The hype would be quite a bit smaller if they weren't calling it "artificial intelligence", which it really isn't.

View this comment - View article - View full comments

Quoting: EikeWhat devs likely would - and will - do is e.g. generating backgrounds instead of painting them. I've already seen this in point and click announcements. Looks very nice on first sight, on second you see some stuff is wrong. If When AI learns, and doesn't paint some nonsense, well, that's easily done and looks nice. And shouldn't infringe anything.

This will probably be a technique that will be used some, but I wonder if it will turn out to work any better than other computer techniques people already use for landscapes that don't call themselves "artificial intelligence", whether it's procedural generation or just hitting real landscapes with filters.

View this comment - View article - View full comments

Quoting: EikeDid you actually try image making AI? I gave it an attempt, and while of course I don't know all the art out there, I'm confident this wasn't just something slightly morphed.

There's a lot of weird stuff on DeviantArt. I wouldn't want to bet too hard if I were you. Certainly when a couple of my friends were fiddling with AI art, they found that when they did variants on some pretty odd requests, there were strong patterns in what they got. Some definitely looked like the same picture was being warped in somewhat different ways.
When I think of it, probably the more "normal" a request, the less of this phenomenon you're likely to see, because there will be more source material and the results will be a blend of more sources.

View this comment - View article - View full comments

Quoting: Eike

Quoting: FSE

Quoting: EikeThey could Paint Yoda and Harry already.

Painting Yoda or Harry is IP infringement unless one can prove fair use.

You missunderstood what I tried to say: They can paint Yoda and Harry already, and thus infringe copyright, just like they can with AI. So I see nothing new with regards to Yoda and Harry.

Yeah, but the point is everything an AI produces is somebody's Yoda or Harry--maybe two or three somebodies mixed together if you're lucky.

View this comment - View article - View full comments

Quoting: Eike

Quoting: hardpenguinYes, this is good. Gotta get rid of the AI generated images (it is difficult to call it art).

It's obvious that this will not happen, right?

I do understand - and share - such feelings, but in the end, it's like trying to get rid of photos in the early stages of photography because they're "not art".

People said similar things about music sharing--where did it go?

View this comment - View article - View full comments

Quoting: Cyril

Quoting: Purple Library Guy

Quoting: CyrilBut the real question maybe is: do you want to learn?

Well, I dunno about Linuxer, but I don't. Or rather, I don't want to learn this topic. There is too much knowledge in the world to learn it all, and I have different priorities.

Even if it means you can't use a software you want?

If it's important enough software I may have to learn that stuff, but I don't want to, it would be hours wasted from my life that I can never get back. I would consider it a failure of the software ecosystem that I was put in that position.

View this comment - View article - View full comments

Quoting: SolarwingI bought Steam Deck last year and I've no desire to buy Asus Rog Ally

Me neither. How has that thing been doing, anyway? We had a big stir about it but since then things have been quiet--are they selling?

View this comment - View article - View full comments