I’ve lost follower acquaintances for saying this, so I’m saying it again: If you think that Statistical Nonsense Machines (wrong terms “AI” or “ML”) make art, you don’t have a clue what art is.

No, really. Go study it.

Art is sociological artifact. Remove the cognition to create interpreted intent in relation to the sign and signified, and the concept of “art” is removed.

Literally *not* art.

@troy_s art is made by human beings, computers ie AI or machine learning can copy the style or content of the art that humans make but if there were no art made by humans there would be nothing for the computers or their algorithms to copy. But yes a sandcastle made o the beach by a kid, or a drawing or a sandcastle by a kid is individual and is art, an AI sandcastle is a copy
@troy_s there may be some computer art databases (I use computer art or machine art because AI implies that the computers are intelligent which is not yet the case) where a mixture of artists and programmers have programmed the computer art generator to produce art in a certain way in response to prompts. In which case the artists are the artists and programmers (involved in developing the generator) not the computer

@Herstory Would probably disagree even from the rather naive “tool” construct. The reason is rather straightforward.

The *logic* behind the sign / signified relationship is critical to the *interpreter*. That is, feeding the Statistical Nonsense Machine a prompt is a fundamental *disconnect* between this critical aspect of art-as-artifact, and mere statistical mashup.

There is a significant cognitive comprehension difference.

They theory of “tool” is predicated on top of nonsense.

@troy_s a brighter side of this would be that artists now will reduce to connect much more through acquaintances to filter through people so the overall quality of the talent pool becomes better 🤔

@troy_s Disagree. These artifacts are created as directed by people and presented by people, which provides the cognition you indicate is absent.

Arguing otherwise is, IMO, arguing in space very much adjacent that which posits the noncogitant paint splatter makes Pollock art not art.

@troy_s I am very happy and eager to entertain discourse that indicates that this tool is harmful in regards to substitution and uncompensated erasure of labor.

@millenomi I am fully amenable to this vantage, *except* in this case. The negotiation of sign to signified implicitly involves the interpreter to close the loop.

Again, definitionally, we cannot displace the *interpreter*, which in this case is not only lacking for sign-signified understanding, but also *outsources* the interpreter as well.

@troy_s I find it incorrect to indicate that an interpreter doesn't exist — there is a person performing a selection of content and a refinement of input (prompts, negatives, weights) to produce desired (thus interpreted and mediated) work. 'Sign' here is IMO used in a purposefully reductive way — if you mean that as a willful single use of a single traditional instrument on a single canvas that is built with significance. Here, the sign is the work produced in its entirety.
@troy_s This is very much 'define a chair' discourse; note how sign-significate reduced this way becomes exclusionary of, e.g. 3D modeling as art.

@millenomi The concept fails on sign and signified in the scope of the mechanics of active interpreter. (At risk of leaning hard into semiotics.)

That is, with a human interacting with a surface, the mark formed is active interpretation. Here, the sign, signified, *and active formulation* of the mark is outsourced.

Could a human form an assemblage of Statistical Nonsense? Absolutely. But the *activity* is the part being missed here. The *activity* is pure nonsense affect.

@troy_s I can entertain that discourse for several use cases of the tool — where a single element is picked out of a production of a single prompt. But the more curation is applied, the less I am happy with this boundary being applied. 3D art fully outsources the active formulation of the mark to hardware, yet we label it as art.

@troy_s I say this with understanding of its devaluation of the labor of the artist, who produced the traits that have been pulled into the model, and for how this is disastrous for the reduction of demand for that labor (e.g. I keep an eye on spaces that use art commercially, like trading card games, and I see an uptake in willingness in producing finished work with AI art over the commission of pieces of traditional art).

I am just really wary of exclusionary definitions of art.

@millenomi Believe me when I say I am not trying to be exclusionary of the notion of "art". I worry that, much like your concern of the labour aspect, that the very definition of "art" has been completely eroded due to lack of investment in the humanities and that insidious war via STEM, but also that there is now a whole culture of folks who are unwilling to simply accept that their understanding of "art" in the broadest sense is that of an atrophied white fish at the bottom of a trench. +

@millenomi - This **is not a fscking accident**! This is an active assault on forms of knowledge and understanding. It's a Foucault-in-a-sound-bite. The entire premise is presented as "Hot Take" because the institutions of capital have *presented* the knowledge structure as such.

TL;DR: The idea that "Everything is art man" is an active force from capital. It's not. And Statistical Nonsense is a symptom.

@troy_s Yeah. I guess I am here to represent the positive aspects of that argument — that if communication is an aspect of the meaning-making of art, we have to pay tribute to the fact that there has been a sea change in communications with a performative or artistic intent — including _primarily noncommercial_ uses of such — by employing recomposition and meaningful curation, memes and fics on up.
@troy_s I understand that this is prompted, encouraged and exploited by capital, as well.

@millenomi I can tell you do. Which is why it's a discourse.

I feel on some level it is a magician trick of capital. The Overton window on the very construct of knowledge, in this case the *definition* of "art", has been grossly hijacked by the historical awareness of a nine year old. And that doesn't give the nine year olds enough credit.

The magician trick here is equivalent to someone reading a Choose Your Own adventure book, and insisting to someone outside, that they have authorship.

@troy_s I feel this discounts the tool entirely. Agency is a gradient and it can produce recognizable authorship, and this is not a CYOA book but it does have enough directive elements (especially if you use the tool directly, not just through pay-to-prompt services) to qualify for sure for more agency.

@millenomi The issue is guidance of the mark. I'm sure you've experimented with even the chat versions... and they too fail in the exact same ways. The guidance of the mark is interpretive loop not only missing here, but the formulation of "meaningful marks" is missing.

Ask how a stochastic diffuse pattern forms "this is meaning" and it reveals the precise faulty foundation of the label of "art".

Most certainly nuance here, and I appreciate that it vibes like a "hot take". Nuance!

@troy_s yeah; I am willing in my definitions here to abstract over the mark as in the specific effect produced on canvas by a specific technique, since this is clearly an accident or second-order effect of the tool.

There's a lot of 'are curation and recomposition art?' and 'how much of either is here?' here as gradients, and I am willing to appreciate how much this tool implicitly pushes people over the shallower end of this pool. Yet even here I am sad to discount the agency of the prompter.

@millenomi Again, you've surely tried it. Have you noticed that there is an almost iron fisted inability to steer the mark? This is not an accident. And this is really exactly the sort of nuance of the interpreter-in-loop between sign and signified here.

And yes, I'd clearly clap for someone who takes the assemblage of clip art of Statistical Nonsense and makes a powerful piece.

But as mark-maker, it is such a shallow construct to as be almost laughable, were it not for confusion.

@troy_s I think the standing point here is that there is an unwillingness to abstract artistic intent out of the medium into an exploration of the possibilities of the tool in itself (including, but beyond prompting), which, I guess, is where we diverge.

@millenomi I'm trying to be *very* specific here. The stochastic statistical nonsense, in the literal definition of "sense" or "meaning", is where I see a critical fault line.

That is, if I asked you "What is the meaning rule that guides the stochastic diffusion of dots toward or away from something?", how would you answer?

I feel this is where the key issue is.

@troy_s I’ve taken a lot of aesthetics and theory classes and this succinctly sums my opinion on what is art, being an artifact of human society! It’s something that happens when we make things. Hard to find the words to describe that so I really like the way you put it.

@emmjemmyjammy And we need to actually focus on the *mechanics* of the creation; the interaction of the human with the formulation.

That is, if we are desperate to take the naive (and false) concept of “tool”, despite it also failing on sign and signification formulation, it *also* fails.

Why?

Because the fundamental mechanic of the *active interpretation*, during the formulation of the marks themselves, is outsourced to mere simulacrum and pure meaningless mimesis.

Okay, but that’s just playing fancily with definitions. In that sense, it’s obviously not art.

But I personally like AI art and can sometimes strongly relate to an art piece created by a piece of code, and/or find it very beautiful…

@troy_s where's the line between algorithmic art and these new things? If the designers of the tool had artistic intentions and an aesthetic they intended to bake into the model, does that make it art? Midjourney seems to have a "house style" that seems very intentional to me.
@troy_s Like, if anything produced by twiddling parameters of a generative tool categorically isn't art then that seems to prove too much
@troy_s eg if I generate a thousand images of perlin noise at random and pick the 10 that look like faces to me and print them out, is that more or less artistic than generating a thousand Midjourney images and picking my favorite ten?

@troy_s my thoughts on this. We have a next-level Turing Test: not "Is this human?" but "Is this creative?"

https://youtu.be/ZyARJ9ZcaNY

Shakespeare and Creative AI

YouTube
@troy_s human exceptionalism at display! As an anti humanist, I completely disagree! Your post is wrong! DELETE IT!

no but this is a very narrow minded view and honestly one mostly held by establishment voices that need something supernatural like their “cognition” as a way to call whatever they make “art” so they can separate themselves from the rest of society

“Art” is like Coca Cola's “the real thing”, just a cute slogan to sell shit

@troy_s here is a brief thought experiment. say someone commissions a painting. the customer is very very specific about what they want it to depict -- objects, their relation to each other, poses, gestures, facial expressions, whatever.

from this description, two paintings are made, one by a human and another by typing the description into one of these programs. is one of these products art, and the other something else?

@troy_s I agree, but do you have another term for it? "Generated images" perhaps?
@thejustkat My preference is for about as successful as unguided diffusion models are at constructing signage... Statistical Nonsense.
@troy_s Why cant the world be run by machines?
@radicalresilience It already is. We call them CEOs that blindly crunch numbers chasing profit.
@troy_s Counter-argument: Current AI systems like Stable Diffusion or MidJourney are just sophisticated tools translating the intend of their users into visual artefacts. Nothing but a really advanced paintbrush cum compass and straightedge with a bit of scissors, glue, and a polaroid in a handy packet. Of course it's art.

@StephanSchulz See the many responses to the naive notion of “tool” here in this thread. Does not qualify as a tool when it outsources the feedback loop between the sign, the signified, and the interpreter.

We ignore the abject disconnect of meaning due to the Statistical Nonsense diffusion pattern, it *fails* on the interpreter front in that the loop of mark making is outsourced to someone’s else’s code driven nonsense of “meaning”.

@troy_s That's not how people I know use these tools. They create multiple prompts, refine both the prompts and the images, do manual touch-up and colour correction, and so on. A similar discussion took place when photography was new - "that's not art, that's just a mechanically produced image". See where that ended...
@StephanSchulz Again, please exert the bare minimum of energy to distinguish what was said and why it is not “tool” class for the reasons mentioned.
@troy_s Well, we may be talking at cross purposes. My point is that "feedback" is not outsourced - it's provided by the user. And of course these are tools. If what they produce is art is under discussion - not there status as tools per se (or we have very different definitions of tool).
@StephanSchulz It *is* outsourced. The marks are not the author, but a statistical kitbash of other marks. Worse than that, the focus of any critique should be upon the *mechanism of meaning*. That is, the outsourcing is nothing but someone else’s idea of “meaning”, inscribed into the algorithm, which guides clusters toward or away. Further still, using these things also reveals this clearly; they are brittle and cannot be coaxed into fine grained constellations. +

@StephanSchulz So I implore you to consider the specifics of the mark making itself, and to understand how wildly different it exists from say, a chisel.

1. The “meaning” directionality is outsourced; the “decision” to move toward or away from a mark.
2. The invisible simulacrum nonsense above, is placed atop of *other authorship* to further distort the mark making.

Reading a randomized set of words as a Choose Your Own Adventure is not authorship.

@troy_s Putting new features atop other authorship is indispensable for all art. Consider Warhol's Soup Cans - he uses existing art and rearranges it. Or john Cage's 4'33 - it only exists in contrast to all formal music the preceded it.

Also consider that many artists use random elements - an extreme case is found art, but Jason Pollok certainly did not plan every individual splatter.

@StephanSchulz Please read what I typed about marks and attempt to understand it. There is a massive difference in this, that folks refuse to see.
@troy_s As I said: We seem to talk at cross purposes. A user of (e.g.) MidJourney creates a prompt - if (s)he has any experience, then that prompt is in a semi-formal language with a lot of explicit instructions, e.g. on style, format, lighting, and so on. That in itself is a creative act. The MidJourney renders several images for the prompt. The user selects one or more for refinement (there is you feedback step!) and iterates the process, sometime modifying the prompt.
@StephanSchulz I assure you I am not speaking at cross purposes. The mark making is the key. But carry on…
@troy_s Well, I'm neither an artist (except in C and Scheme) nor an art specialist, nor even a native English speaker. But "making the mark" (which I understand as the physical act of modifying the surface of an artwork) seems to be a trivial, merely mechanical act. Many artist use virtual tools with pre-defined brushes and textures - is that not art for you?
@StephanSchulz Focus deeper on the *guidance* of the art. That is, what drives meaning of the actual interpretation and selection of the mark, at the lowest level.
@troy_s But that is exactly what the user is doing - it guides the system to his desired result. He's doing it at a higher level than with a brush and paint - but then so is a photographer.
@StephanSchulz I’m sorry. The very first post hinted at this distinction in terms of interpreter relation to the guiding of the mark. I cannot force anyone to focus on the actual thing I just described. Note how you avoided explaining what I asked you to explain?
@troy_s I suspect I don't understand what you want me to explain. And I also think our discussion lacks a clear definition of the term "art". Can you imagine a computer ever to create art? Can you imagine hypothetical intelligent aliens to create art?
@StephanSchulz Try to understand the thing I asked you to explain.
@troy_s I think we must distinguish different things. First, ist the output of MidJourney art? Secondly, if so, who is the author? The user? The AI? The programmers? The authors of the art that was used to train the AI? Any combination of the above?

@troy_s That is a good point, what people want when they say "art" these days would be better termed "illustrations".

(Like, every RPG these days is expected to be filled with illustrations)

@troy_s I totally agree.

Machines can never make art without an underlying understanding and desire to convey message and meaning.

But an artist can use a tool to create art.

What if the artist replaces a paint brush with a computer? Does it cease to be art?

I don’t have any good answers, but your message spurred some intellectual curiosity in me.

@anders I think these are all valid questions.

On the subject of “tool” it fails miserably as well. The “meaning” is driven by stochastic diffusion guidance, and as such, the interpreter-sign-signified loop is (a failure of course on even sign and signified) is *snapped* via the outsourcing of the interaction between the interpreter and the mark.

To be clear, a human can create a powerful piece via assemblages of Nonsense, but the singular piece *does not qualify*.