scoobdog

Indeed, but can any of that truly make it feel an emotion?

The question isn’t whether it knows how to portray an emotion or even elicit that emotion. The question is whether it can have its own emotional interpretation of what you provide it. You’ve already pointed out that the person giving the prompts isn’t an artist, but if not, who is? Art still has to created by an artist. Can the AI be considered an artist if it’s perspective is incomplete?

Given his perchance for showing up unannounced at the most inopportune times, I would think any radio silence from him would be as great a cause of concern if not more than if he was actually missing.

How do you do that? Let's assume that you're not just providing the AI with prompts, you're also limiting its sources by only allowing it to use photographs you took yourself as primary inspiration (for POV, base colors, and peripheral landscape elements) and curating related images that can only be used as indirect secondary inspiration (color juxtapositions and degree of focus). How are the choices it makes to interpret your vision also conveying emotional elements you can't express in prompt form?

That's an obvious case of artistic dishonesty. That also leads to the question of "How can AI generated images become art?" Sponges has, to this point, not really touched on the actual definition of art. For most people the Potter Stewart definition is sufficient, but AI forces us to reckon with art as a social implement, both as a means of conveyance and as a form of currency. How much investment the artist puts into the art directly correlates to how much value it has as a conveyance but only partially correlates to its value as currency. That brings us to our second point of divergence: how much does the AI generated image's value as a conveyance determine its status as art? If Shexyo is selling retouched AI work, the theft isn't so much in the fact he used AI, it's in the fact that he failed to disclose his use of AI. The fundamental disconnect here is between those who purchase art and the artists themselves, namely how each defines art. If it's a commodity, then provenance is part of how an image is appraised along with its aesthetic qualities and the difficulty with which it can be duplicated (as a genre piece, that is). Yet, art isn't pursued as commodity explicitly. If you trace the career paths of some of the artists that were most successful in their own lifetime, you notice that nominal value for their work doesn't correlate to the actual value. A lot of them succeed through ancillary factors, like a patronage with a wealthy client, lucrative speaking engagements, or other fame related benefits, meaning there's a distinct difference between the cost of replacement for the art and the cost of replacement for the artist. Such a distinction is important because it suggests that you can't use the art's value as a unique piece, an element of currency, to define it as art because the cost of replacement varies throughout the lifetime of the artwork and art itself is a static definition. In this respect, Sponges is also right. However, if it still holds true that investment correlates into value as a conveyance, then it stands to reason that its value as standalone art also correlates with that investment. We don't just define art by its final product, we also include the various elements that comprise the completed work as having value. In traditional art that might include the brushstrokes and peculiar pigment choices of the artist, but in digital art, that means more of how component elements are juxtapostioned and then "built upon" to create the final image. (A familiar artistic trope is that no piece of art is ever truly complete, in part because each piece reflects not just on it's predecessors but provides a blueprint to future evolutions in the artist's techniques.) With AI, the investment is mostly on the part of the computer because (as mentioned in a previous post) the only thing it can't perform itself is to "see" what needs to be reinterpreted into the new unique image. Just as with a human artist, the AI "artist" is using past images not just to inspire its work but also to develop its technique and, also like a human, it can't create a final product and still learn from the process. It's hard to overstate how contentious this point is, though: by failing to see and emotionally interpret the inputs that are used to create the image, it's missing a very large piece of investment. AI might be able to manipulate how the image is received by learning how to adjust perspective or how to use color that elicits an emotional response, but it's not sentient... it has no reaction to the images it uses. At the very least, that suggests its worth as a language is limited by what it can't provide.

I think it's about Sponges himself. He likes to make things even more complicated.

There is room for nuance, but the AI isn’t critiquing your work, it’s providing an alternate vision to yours. That’s fine when you’re using human artists as a gauge and inspiration. An AI, however, is devoid of emotion, leaving just its (limited) technique as a source of collaboration. How much you can derive from its product remains to be seen in such a limited scope.

You can't script a result based solely on prompts, though. At least in that respect, Sponges is right - there is a learning curve involved in these programs that is greater than the sum of just the inputs and the results. I would also argue that it isn't a valuable learning tool because of the fact it has the ability to learn.

I assure you it's not a gotcha. I'm really trying to identify point of divergence in this argument because this is more complicated an argument than just theft versus coopting. Let's start with this: So, you're out for a drive along the Lake Michigan coast and you see a duck floating along as you drive. The duck isn't uniquely native to the lake, nor is it out of place. Let's say it's a fairly common Mallard. Now, after you get back home, you decide to paint the duck you just saw, so you get out canvas and start drawing a basic sketch you can will use as an outline for your painting. You don't much think of it while you're drawing but, as soon as you're done, you think for a moment.... "Is this the duck I saw or did I copy it from something else I've seen?" The simple answer is, of course, that it can't be a copy... you saw the duck on the lake; it's a common duck that a lot of people have seen; and there are only so many ways to interpret a duck that are unique. There is commonality in our view of the world that we share with all humans, so what is or isn't a copy is almost always determined by whether or not you have an identifiable source. Supposing another artist comes up, looks at your paining and says "that looks almost exactly like mine," they would have to prove that you not only have seen their painting but that it's more likely that your painting could only be produced through seeing that painting and not by a common experience like taking a drive along the lake coast. But, assuming you've seen that other painting, it’s also true that it had some kind of influence on your work, regardless of whether or not you thought about it when you were painting yours. This, then is the first point: you process other people's images like the AI does, but you also have the ability to see it yourself. Can the AI, likewise "see" an image not produced by someone else and posted in the internet?

To put this disagreement to bed. You obviously are talking about something far different than everyone else is talking about. If you can generate an AI piece, then you can use it to illustrate the process that you’ve been describing.

You had to see it. MTU intercepted the Aztecs on SDSU’s 25, then they proceeded to give up four straight sacks and ended up giving the Aztecs the ball back on downs on the Raider’s 45.

With some in just t-shirts.

Alright dude. Create some AI work for us and post its here.

It’s piecing together this “shit that’s all fucked up” from other people’s work. The AI didn’t take draw an original image of Donald Trump from a photo, it found drawings of Trump and incorporated them.

We wouldn’t be having this discussion if the pattern seeking components didn’t generate images that looked almost exactly like existing images.

You came off as abrasive? The fact is, to a person every opinion against has focused on the theft of existing images being the central issue against both AI art as truly "art" and AI as being a viable medium for that art. To be perfectly clear, AI artists are "lazy" because they're utilizing existing images created by someone {or something) else, and that is a completely valid complaint. You can't grab an image, modify it, and repurpose it as your own work - that is theft. Warhol pushed the boundary by utilizing "popular" images in mostly recognizable forms, but his work is displayed in art museums because it (1) modifies those images into their deconstructed parts and (2) is direct commentary on these images as part of popular culture (hence "Pop Art"). There is a lot of back-end and off-canvas work that goes into repurposing these images into high art, and that work reflects a specific framework in which that art can be viewed and appraised.

That is an excellent point: framing it not as a feds versus states but as states versus states does change things considerably.

You would think. There's a whole lot to lose when the inevitable detente between the feds and the legalized states is broken.

I think it's simpler than that. Politicians across the spectrum have other priorities and they're not motivated either to expend their political capital to get it passed or have no particular reason to budge on opposition. The fact that we've gone as long as we have with state legalized pot with no real interference by the feds leads a lot of legislators to mistakenly think they can continue to punt the issue.

I think @NewBluntsworth might be interested, but you’ll have to PM him I think.

I am not. I didn't say it wasn't art, I just said it served no real purpose. You serve no purpose. Sorry that was childish of me. But seriously, that argument is ridiculous. How can it serve no purpose? Conveying meaning, evoking emotion, aesthetic beauty in and of itself? Advertising by ghoulish corporations? Because it's used for all of those things, which inherently means those things be their purposes by transitive property of fuck all. (In FESTIVE GREEN!!!) See below, the image doesn't convey anything because it's just an image created by a third party. Even if I know what your prompts are, how do I know what your intent at the start was and what is the computer's choice? Art isn't just an image, it's an image the artist wants his viewers to see. Art as a form of expression ultimately is defined not by what it says but who it says it to. Art isn't defined by what it says. But who it says it to. What. Art is absolutely defined by what it says, I mean, how could it not be? And yes, the interpretation of the audience absolutely helps define what the art says. Art isn't defined by what it says: Hey boss, drew this pineapple. No reason. But who it says it to: I am your boss and I am moved by that pineapple. I said what I said. As a writer, i understand art as having a fundamental barrier. I can tell you a story and use the most precise language my skill allows, but there will always be a point where my interpretation of a character's words, actions, and intents are different than that of my readers. The same applies to visual art: I can paint you a forest scene at sunset, but what that scene means to you will never be exactly what it means to me. Therefore, art is created with the fundamental understanding you aren't manipulating your audience, you're giving them something you reasonably expect them to interpret a certain way. Prompting an AI to generate a clown Trump out of stolen drawings might certainly be a mirror to your own wants and desires and as a form of art that has some inherent value, but it says nothing to other people because it doesn't include any of your intellectual framework. It's just an image that you told an AI to create without any context. To Buddy's point... You're being intellectually dishonest comparing the lowest common denominator-appealing images instead of things that actual thought has gone into generating. My drawing a weiner on a piece of paper with a pen does not denigrate the medium. More than just expression, art is a form of language, or more accurately, a class of languages. Structuring this in terms of accomplishments and results tends to obscure the proper intent; there is, nonetheless, a component of effort that must be factored. DaVinci is the gold standard in this metric because he is almost always lauded for his technique rather than his subject matter. We don't have records of his artistic process either directly from him or by eyewitnesses, but we can actually see the brush strokes on his canvasses and see the choices of pigment for the paint in certain places. Now, with the use of x-ray, we can even see the evolution the painting, so we can actually piece together the progression of the painting from an earlier draft to a final piece. To make a somewhat hackneyed comparison, each brush stroke, scratch, color change is a different "word" in the collective "novel" that is the completed painting because those are individual parts that can be interpreted and directly contribute to the final piece. How much human effort into alteration of an AI generated work would allow the piece to be considered legitimate artwork? Someone generates a concept for an image, opens it in photoshop, and edits it until they're happy. How much editing is the threshold? Shit, how about someone creates their own original image, uploads it, and edits it slightly using the AI? Or edit it a lot using the AI? Because you can do that too. If you: (1) directy input the images used by he AI, (2) prompt the AI how to interpret those images, and (3) can explain the process the AI used to generate the result (as in can explain the AI's process for it) you will have created a legitimate work of art. Condition one almost certainly mandates you create your own parent images, but you very well could use images with the permission of their original creator / owner (in the manner of, say, Andy Warhol). You still need to tightly control what images are used and have an understanding before you use them on how they will be used. Like Buddy says, the person who inputs the prompts isn't creating anything. The tool isn't a dumb implement like a brush or a paint palette, it has the ability to make choices on its own, and, consequently, there are no brush strokes or pigments to analyze. The artist has no say other than the inputs he or she puts into the AI. He or she doesn't even have the ability to select which images can be used by the AI to create the amalgamation. As the tool is currently structured, the AI results can't be considered true art because the artist doesn't make all the choices to create the final work. I think there's already a spectrum of degrees of artistic control when it comes to legitimate work. Many artists have utilized a lack of control when it comes to their medium. Droplets of paint raining onto a canvas doesn't require the artist to make all the choices to create the final work. No artist truly lacks control of his own work. He may choose to use random variables as a part of the process, but he still has to have a final vision that the work will ultimately meet.

Have you already exchanged gifts? I only said what I said because I haven't heard anything on my end so I assumed nothing has happened yet.

I am not. I didn't say it wasn't art, I just said it served no real purpose. Art as a form of expression ultimately is defined not by what it says but who it says it to. Prompting an AI to generate a clown Trump out of stolen drawings might certainly be a mirror to your own wants and desires and as a form of art that has some inherent value, but it says nothing to other people because it doesn't include any of your intellectual framework. It's just an image that you told an AI to create without any context. To Buddy's point... More than just expression, art is a form of language, or more accurately, a class of languages. Structuring this in terms of accomplishments and results tends to obscure the proper intent; there is, nonetheless, a component of effort that must be factored. DaVinci is the gold standard in this metric because he is almost always lauded for his technique rather than his subject matter. We don't have records of his artistic process either directly from him or by eyewitnesses, but we can actually see the brush strokes on his canvasses and see the choices of pigment for the paint in certain places. Now, with the use of x-ray, we can even see the evolution the painting, so we can actually piece together the progression of the painting from an earlier draft to a final piece. To make a somewhat hackneyed comparison, each brush stroke, scratch, color change is a different "word" in the collective "novel" that is the completed painting because those are individual parts that can be interpreted and directly contribute to the final piece. Like Buddy says, the person who inputs the prompts isn't creating anything. The tool isn't a dumb implement like a brush or a paint palette, it has the ability to make choices on its own, and, consequently, there are no brush strokes or pigments to analyze. The artist has no say other than the inputs he or she puts into the AI. He or she doesn't even have the ability to select which images can be used by the AI to create the amalgamation. As the tool is currently structured, the AI results can't be considered true art because the artist doesn't make all the choices to create the final work.

My sympathies

How do keep your job?