Because it is mimicking human input. Effectively you are getting a mixture of many pieces of artwork that humans made distilled down into some sloppy new one that was made without feeling, purpose or skill and that can be described by its prompt, a few kilobytes at best. Original human art can only be approximated but never captured with 100% fidelity regardless of the bitrate, that is what makes it unique to begin with. Even an imitation by another human (some of which can be very good) could stir you in the exact same way but they'd be copies, not original works.

Anyway, this gets hairy quickly, that's why I chose to illustrate with a crappy recording of a magnificent piece that still captures that feeling - for me - whereas many others would likely disagree. Art is made by its creator because they want to and because they can, not because they are regurgitating output based on a multitude of inputs and a prompt.

Paint me a Sistine Chapel is going to yield different results no matter how many times you would give that same prompt to Michelangelo depending on his mood, what happened recently, what he ate and his health as well as the season. That AI will produce the same result over and over again from the same prompt. It is a mechanistic transformation, not an original work, it reduces the input, it does not expand on it, it does not add its own feelings to it.

I think this is a reasonable counter in some respects, although I do also think it's specific to the current iteration of AI art.

It's a bit like when people describe how models don't have a will or the likes. Of course they don't, "they" are basically frozen in time. Training is way slower than inference, and even inference is often slower than "realtime". It just doesn't work that way from the get-go. They're also simply not very good - hence why they're being fed curated data.

In that sense, and considering history, I can definitely see why it would (and should?) be considered differently. Not sure this is what you meant, but this is an interesting lens, so thanks for this.

Haven't these arguments been the same since Stable Diffusion came out? Someone (A) will say what you said, then someone else (B) will say, well humans remix as well, A: no that's different because we're humans not machines, B: there is no need to prefer a biological substrate over a silicon one; A: AI will produce the same result over and over, B: not if you change the temperature and randomize the seed.

It's tiresome to read the same thing over and over again and at this point I don't think A's arguments will convince B and vice versa because both come from different initial input conditions in their thought processes. It's like trying to dig two parallel tunnels through a mountain from different heights and thinking they'll converge.

Ironically, for the first time, I think I found some perspective to the remix argument here.

Normally it's just like you say: I don't find the remixing argument persuasive, because I consider it to be a point of commonality. This time however, my focus shifted a bit. I considered the difference in "source set".

To be more specific, it kind of dawned on me how peculiar it is to engage in creating art as a human given how a human life looks like. How different the "setup" is between a baby just kind of existing and taking in everything, which for the most part means supremely mundane, not at all artful or aesthetic experiences, and between an AI model being trained on things people uploaded. It will also have a lot of dull, irrelevant stuff, but not nearly in the same way or in the same amount, hitting at the same registers.

I still think it's a bit of a bird vs plane comparison, but then that is also what they are saying in a way. That it is a bird and a plane, not a bird and a bird. I do still take issue with refusing to call the result flight though, I think.

Flight has immediate utility, art not necessarily, other than to be or to experience. Movies can be art, instruction videos usually are not.

Flight isn't necessarily utilitarian. Not animals', not machines'.

A connected discourse is (certain, increasingly dwindling maybe) part of the art community's rejection of large swaths of works because they're meant for mass entertainment.

And so I'm not sure robbing AI generated images of being labeled art isn't a similar kind of snobbery, at least in part, with models just being a much more morally convenient punching bag this time around than other humans.

Something not being necessarily utilitarian does not mean that it isn't mainly utilitarian. There is knitting as an art form. But it was definitely mainly utilitarian at some point.

And this is how it goes with many things: at first we do them because they are utilitarian, after that there may be people who start using it as a medium for art.

> And so I'm not sure robbing AI generated images of being labeled art isn't a similar kind of snobbery, at least in part, with models just being a much more morally convenient punching bag this time around than other humans.

Then show me the art. Just one single image that moves you and that was generated by AI.

> Something not being necessarily utilitarian does not mean that it isn't mainly utilitarian.

In terms of extents, I'd say machine flight is about as utilitarian as animal flight. Which is why you don't see it differentiated in verbiage I'd imagine. I'm generally not sure where you were going with this.

> Then show me the art. Just one single image that moves you and that was generated by AI.

There isn't a single drawing (picture) that I remember to have ever moved me, manmade or machine generated, so that's quite the tall order.

For examples on AI generated images I see, that'd be on Pixiv. They're almost always tagged up and you can filter for (and against) them. And there are of course people who exploit this for harassment, because no good deed goes unpunished.

With the proliferation of AI, I saw styles, poses, framings that I haven't before there, as well as their combinations. Were they just underrepresented among other people's drawings? I'm not so sure - some are for sure referencing actual photographs instead, and some are assisted rather than fully generated. I did enjoy these greatly, even though they were not straight from the remotest figment of someone's personal imagination, and they haven't per-se "moved" me.

Ok. Thank you for the answer and the exchange in general. I suspect one part of the issue here is that some people are more sensitive to stuff like this than others.

For instance:

https://www.youtube.com/watch?v=o_wsSIuv_po

Never fails to give me gooseflesh every time I listen to it. And where it gets interesting is that that is a cover of a piece by another composer, so it serves as a very high level commentary and compliment rather than an original and still manages to maintain a lot of the emotional content and adds new elements. The original is:

https://www.youtube.com/watch?v=vE2O_yfgtBU

Adagio starts at 3:32.

See if you get a different take away from each. I find both beautiful but as different as jam and cheese.

There are drawings and paintings that move me in a similar way. And I'm sure there are people who are not touched by any of this. I've been steeped in art pretty much since I was a toddler, my dad was a painter (in my opinion not a very good one but that did not stop him from endlessly trying) and our house was always full of music, antiques and conversations about that stuff. This probably sensitized me in a way that I would not have been if not for that environment.

The interesting thing is: even bad art is still art.

The day I see AI generated art and it moves me in the same way that human generated art does I will concede the point. So far all I've seen is more, not novel.

Art never was about productivity, even though there have been some incredibly productive artists.

Some of the artists that I've known were capable of capturing the essence of the subject they were drawing or painting in a few very crude lines and I highly doubt that an AI given a view would be able to do that in a way that it resonated. And that resonance is what it is all about for me, the fact that briefly there is an emotional channel between the artist and you, the receiver. With AI generated content there is no emotion on the sending side, so how could you experience that feeling in a genuine way?

To me AI art is distortion of art, not new art. It's like listening to multiple pieces of music at the same time, each with a different level of presence, out of tune and without any overarching message. It can even look skilled (skill is easy to imitate, emotion is not).

I still don't get why you don't see it as a tool and not the creator itself. The human sitting behind the desk is the one attaching their emotions to what they send, because they control what image they want to send, otherwise they reroll or redo their work flow. These days they can even edit the image with natural language so they can build it up just as one does in Photoshop, only using words instead of a mouse.

> I still don't get why you don't see it as a tool and not the creator itself.

If after 33 comments in this thread and countless people trying to explain a part of it you don't get it that may be because you either don't want to get it or are unable to get it. Restating it one more time is not going to make a difference and I'm perfectly ok with you not 'getting it', so don't worry about it.

AI without real art as input is noise. It doesn't get any more concrete than that. Humans without any education at all and just mud and sticks for tools will spontaneously create art.

Or perhaps your initial premise ("AI without real art as input is noise") is simply wrong. By "get it," I'm trying to understand why you'd believe such a premise, yes even after 33 comments, because there is no underlying rationale to it, or rather, you never state it in a direct manner.

This is where you might be "not getting it". A human can carefully weigh every word, every swipe of a brush, or every tone... weigh it for the emotional expression and connection it produces (frequently subconscious). Whereas AI as a tool simply can't.

This is a difference between using a gradient in Photoshop, which is still a tool, and generative AI which will make "decisions" you as an author can't explain or connect with.

How is this different from an electronic music producer? They similarly arrange notes without having played them physically. So too with people generating an image as a rough draft then editing every part of it, which is mainly what I'm talking about, not someone who types in a prompt and accepts whatever comes out.

Some people are simply irrational, and there's no point trying to point out to them their logic errors.

>> The human sitting behind the desk is the one attaching their emotions to what they send

natural question: to you draw? Even a simple thing, even a doodle of a cat would count. A particular emoji drawn for a joke. Have you ever drew a line, and then smile to yourself "yes, that is what i want other people to see?"

People can draw poorly or make collages, and come up with pretty expressive art. Those who say "well I can't express myself with stick figures" coincidentally can't express anything without stickfigures too. They just never payed enough attention to the subject to express it.

Personal anecdote: when I ask people why X is in the art they send me, they answer happily. When I ask people with AI art that, they say "oh, you nitpicking". As if some details don't and shouldn't influence art expression. As if all details that weren't in a prompt, shouldn't express anything.

AI art is a concept muddled. It's a grave for intentionallity. It's not easy to decipher creators intent through a cacophony of other intents mixed in because almost none of art choices were made with the intent to convey.

There is no intention in either case. Just a machine doing machine things.

The intention is the human prompting or creating the work flow, the computer was never going to autonomous create images, why would it?

Don't also forget:

A: but AI only interpolates between training points, it can't extrapolate to anything new.

B: sure it can, d'uh.

It's not. If one takes the fact that art is in the eye of the beholder [0], then yes, even AI art may stir you, especially as a human is the one generating at the end of the day, for a specific purpose and statement about what they want to convey.

There is a good part of the series Remembrance of Earth's Past (of which The Three Body Problem is the first book) where the aliens are creating art and it shocks people to learn that the art they're so moved by was actually created by non-humans. This is exactly what this situation with AI feels like, and not even to the same extent because again AI is not autonomously making images, it's still a human at the end of the day picking what to prompt.

[0] https://en.wikipedia.org/wiki/The_Death_of_the_Author

> it's still a human at the end of the day picking what to prompt

I think that 'dutch people skating on a lake' or 'girl with a pearl earring' or 'dutch religious couple in front of their barn' without having an AI trained on various works will produce just noise. And if those particular works (you know the ones, right?) were not part of the input then the AI would never produce anything looking like the original, no matter how specific you made the prompt. It takes human input to animate it, and even then what it produces to me does not look original whereas any five year old is able to produce entirely original works of art, none of which can be reduced to a prompt.

Prompts are instructions, they are settings on a mixer, they are not the music produced by the artists at the microphones.

Have you actually used image generators today? It can produce things it's never seen if only you describe the constituent pieces. Prompts are a compressed version of the image one wants to create, and these days you don't even need "prompts" per se, you can say, make a woman looking towards the viewer, now add a pearl earing, now adjust this and that etc.

> Have you actually used image generators today?

Why would you ask this? It sounds like a lead-up to some kind of put down.

> It can produce things it's never seen if only you describe the constituent pieces.

It can produce things it's never seen based on lots of things that it has seen.

> Prompts are a compressed version of the image one wants to create

They emphatically are not. They are instructions to a tool on what relative importance to assign to all of the templates that it was trained on. But it doesn't understand the output image any more than it understood any of the input images. There is no context available to it in the purest sense of the word. It has no emotion to express because it doesn't have emotions in the first place.

> and these days you don't even need "prompts" per se, you can say, make a woman looking towards the viewer, now add a pearl earing, now adjust this and that etc.

That's just a different path to building up the same prompt. It doesn't suddenly cause the AI to use red for a dress because it thinks it is a nice counterpoint to a flower in a different part of the image because it does not think at all.

I think you're reading too much into my comment. It's not a put down, I'm genuinely asking because it seems many people still think anyone serious about AI just types prompts into Midjourney, but it's become a lot more complex than that, akin to electronic music production; producers haven't played every single note with a physical instrument their synths synthesize yet their arrangement of the notes is what makes them a producer, and so too with AI workflows such as those seen in ComfyUI. If one is not familiar then they might not understand where the field is today.

Regarding prompts, I never said a computer "understands" or is "emotional" about an image, I don't think anyone actually thinks that, on either side of the debate so not sure why you're bringing that up. By "compressed" I just meant in the information theory way, in that if you have a specific series of words, and a given temperature and other settings for a given model, it will deterministically produce the same image, hence the set of those attributes can be thought of as a compressed representation of that image. I made no claims about it thinking whatsoever.

> It can produce things it's never seen based on lots of things that it has seen.

Yes, just like humans, as I had said in my initial comment about the same old arguments being said since 2021 when Stable Diffusion came out. But again that's tiresome so let's not repeat that here too.