[zamadatix]

The output you create with tools doesn't have to be any less rewarding that the output you create without them. A good game isn't inherently meaningless to your life because you used existing tools far beyond your capacity to construct alone (a computer, a compiler, a game engine) and a bad game isn't inherently meaningful to you just because you made it out of a tree with your bare hands. How you wield the tools (innate or external) is much more relevant to creating something meaningful to you and should extend well past when we have tools "bettet" at doing something than we are ourselves.

[chefandy]

Not a popular opinion around here, but generative AI has little overlap with the processes common to creating all sorts of art, but lots in common with commissioning creative work. When you commission art, you have a discussion with the artist— however long it needs to be to get your point across. That can be quite lengthy and involve many sketches, mood boards, etc for important pieces, or a couple of sentences for others— and there are usually iterations where things get modified based on your input. When I do visual design work, I build those two things into every contract I put together. Using generative AI uses almost exactly the same process as commissioning art — you're just commissioning it from a computer that will give you an amalgam of other people's art much more quickly than humans can produce it. That's fundamentally philosophically different than using a tool that will make it easier to put elements that you made into a piece exactly as you want them. Someone compositing with photoshop or doing layout with indesign or whatnot can't call themselves a paste-up person or a dark room photo editor, but they are still deliberately placing elements they made to look exactly like they want, from their own minds, into a piece that they create. They are creating art. I've seen few people using generative AI in my particular disciplines that even vaguely understand how many decisions that algorithm is making for them in that process. They just don't understand it enough to know what they don't know. That's the point, right? Nothing wrong with that, but it's distinct from creating art. It's not even bumper bowling— it's directing an amalgam of all bowlers and taking credit when it gets a strike.

If you commissioned work from an artist, you wouldn't simply cross out that artist's name and add your own, would you? Art directors and editors who exercise precise control over their artist's output don't remove the artists name and add their own, do they?

If there was an AI system that could competently perform surgery, no matter how much massaging that prompt needed, I wouldn't call myself a surgeon for using it, and it would a fundamentally philosophically different activity than a surgeon performing laproscopic survey, evergreen though they're assisted by a modern tool.

[zamadatix]

I agree if you just say "make me a painting" you're not an artist, "write me a program that" and you're not suddenly a programmer, and "operate on the patient" you're not a magically a surgeon. At the same time, if, without AI for a second, you just drag the color pencil across some paper a few times, write int main(){return 0}, or chop the guys pinky off with an scalpel to fix a wart you've not demonstrating being an artist, programmer, or surgeon just because you used the tools artists, programmers, and surgeons are known to use.

There are ways to use AI, or any tool we've yet made, like you're saying some people use e.g. photoshop. I'd be surprised if you've seen most people who own a copy of photoshop learn how the masking, convoluting, and effect layers work algorithmically as their path to learn how to create art with it. Most learn by seeing what can be done with it, mimicking that until they understand how to use it to do what they have in mind, then applying that process. Just as you can deliberately work with traditional tools to piece together and grow an image, you can do the same with generative and diffusion tools as well - their usage goes far beyond "make me painting" in the same way the usage of a brush goes far beyond being able to make the canvas brown instead of white. I don't need to know how the magic wand makes feathered selections of pixels to understand I can make it select sections of an image for me and I don't need to understand how an AI model knows what part of the image was the cats head to have it modify it. I'd also say the vast majority of people with a paintbrush or copy of photoshop have a clue how to make art like a professional artist would, that says nothing about whether one can do so with said tools. It is indeed much easier to create something that used to take a lot of effort, but that's not the right measures for whether something can be used in creation or whether something is good art rather a measure of whether something had any real thought or effort put into it.

I'm not quite sure what you mean about crossing names out, unless you mean what I described above where you say "make me painting" and act like you're a great artist because your tool made something appear instead of you creating something with it. If there is another meaning to what you mean hear it'd be interesting to hear precisely what you mean.

That is to say, there is more to whether doing meaningful creation is still possible than rushing to see whether a similar seeming end result can be created with the same tools in a less meaningful way as well. You wouldn't say Van Gogh's paintings were all meaningless fake art creations because an AI tool can output something which seems like them by simply asking it to so why is it this is the evidence there is no way to use an AI tool to truly "create" any kind of art at all?

[chefandy]

Firstly, the entire premise of what I said is that creating AI images does take effort and creativity; I never once equated it to saying "make me a painting." But mere effort and creativity aren't the baseline for saying you're making art.

Let me describe a process: "You describe a piece of art using words and optionally inspiration pics to an entity, and embark on an iterative process where it gives you images, and you guide the entity with words until it yields a piece you are satisfied with. While you are directing the entity's efforts, ultimately, the work produced was generated by the entity, from it's own experiences and artistic sensibilities."

That entity could be a commissioned artist, or an AI image generator. There is no reality in which that entity could be a paint brush, Adobe illustrator, or an iPad drawing app. When you commission art, you don't say "make me a picture of X doing Y" and see what they make... There's a back-and-forth, preliminary sketches, iterations and adjustments along the way... Just like you use generative AI. It's just what it is. It's completely distinct from any form of creating art. You are commissioning art from a computer or a person. It's not even like it's always less creative. Art directors often have way more of a creative vision in what they're doing then the artists do— but the artist is still credited with making that piece because even though someone told them exactly what to make, they still made it.

I make algorithmic art. In fact, I make art using nearly every level of technological involvement possible, both professionally and as a hobby— from charcoal drawings on rough paper, to code-generated art using Adobe products, Houdini and other 3D applications, shaders, and all manor of other methods. When I am using code to generate algorithmic art, the product is likely even more deliberate and controlled, and the results even more intentional than with charcoal.

I also use Midjourney to generate mood boards, inspo pics, and references. It's great for that. I can get an idea of how something would look without scouring image sources, stock photos, etc. Sometimes I have to put quite a bit of effort into getting the exact angle or texture I want to see, and it's sometimes not up to the task. This is a creative, challenging process, but what it yields is not my art. It's an amalgam of other people's art assembled by a computer program to my approximate specifications.

> I'm not quite sure what you mean about crossing names out

I mean that no— using Midjourney et al to generate images from other people's art is not making art. At best, you could argue that Midjourney is the artist that you are commissioning art from. And just as it would be wrong to put your own name on work you commisioned from an artist, it's wrong to put your own name on work you commisioned from a computer.

> You wouldn't say Van Gogh's paintings were all meaningless fake art creations because an AI tool can output something which seems like them by simply asking it to so why is it this is the evidence there is no way to use an AI tool to truly "create" any kind of art at all?

I don't even know what you're trying to imply I'm arguing here. No artist's art is affected by someone else ripping it off. AI tools do have a place in actually creating art. Even using the images directly, perhaps, by directly manipulating them to make a collage. But no— no matter how many iterations, using whatever long string of bizarre references and terminology you like, you are still asking a computer to make art for you. The most basic fundamentals of composition and construction are the least controllable aspects of this process. No matter how specifically you reference styles or seemingly incongruous terms to trigger some kind of output, the generator is still making the overwhelming majority of the artistic decisions. I use these tools professionally — I know the difference. There are plenty of professional generative AI tools that aren't based on using words to ask it to make you something. Screen tracking and other tools in higher-end compositing software often uses generative AI. You're still executing your exact intention created by you just not having to do one technical part by hand. That doesn't exclude it from being art. The second you start asking something to make something for you, sorry. Nope.

Your essentially arguing that you should be able to claim credit for a book generated by your chatgpt prompts because professional writers use spelling and grammar checks.

[zamadatix]

The premise here was never art either, it was creating and making things meaningful, but that's okay because I'd rather talk about the topic than at each other but semantics on what each of us typed last comment. I think some "you"s which were meant as the general "a person" instead of "chefandy" may have been conflated too. The joys of plain text exchange :), I'll try to use alternative wording on that which is more clear.

In general, I disagree with the idea that because one guides the tool with words descriptively one is no longer able to lay a claim they create something themselves with it. That it can be a way to guide a commissioned artist does not direct whether using that method is always like just commissioning something. First, on the idea of guidance being a universal decider. If one used an intermediate person who could be a commissioned artist to draw a pointillism style piece by giving them 10,000 commands like "3 dot widths left dot again" the creation is clearly still there's despite being able to instead commission a pointillism piece instead. This is obviously an extreme example and clearly commissioning art is actually a thing at some level of command, but the idea is the existence of base commands with something that could take commissions isn't the decider of that. Moving further up abstraction to the second idea, the idea of descriptive things being a universal decider. If one instead says "left, no a little farther left, dot there" and the artist chooses a darker red have one has now lost any claim to creating the output? What if one drags a brightness slider down or tell "dumb" (i.e. non-"AI") software over voice "lower the brightness slider some" and it picks a value until the person is happy? It doesn't make sense the "dumb" slider software suddenly became an artist that made the painting nor does it make sense that using a tool which could make art on its own that way means high level commands like "move the cat slightly left" are proof the work is now commissioned as one didn't move the object themselves. Arguably even if one says the intermediate did become part of creation just on the basis their a "smart" tool it's still not clear why that means one must have no claim to have done any meaningful creation at all with sole claim going to the intermediate now.

Does this mean it's exactly the same as using a brush? No, absolutely not. Your example of someone with a Photoshop not being able to claim they are a dark room editor was a great example of this. Unlike a dark room editor, being an artist or, more generally, a meaningful creator isn't defined by which tools one has used and how one has used them. "Artist", while a varied definition, is surely a much more generic term about creating things than that. If one uses "AI" tools or even normal "dumb" applications like Photoshop one is definitely not a master painter but that doesn't tell us why one is not an artist. Does this mean any form of describing an image is meaningful creation of art? I don't think so, that's what "make me a painting" is supposed to describe. But, again, just because you can say "make me a painting" it does not mean the only thing the tool can do is create things itself. The question is and has been about when something is no longer your creation not whether the tool can also create things where the vast majority of the creation occurs in it. When it comes to whether using an AI tool is for crossing names out I think it is no different than saying your positions on the above. If one conclude along one persuasion it's so, if one concludes along another it's not. I don't think it therefore provides extra insight into what kind of creation is going on but I could well be wrong on that account.

As to whether I'm saying you should be able to claim credit for a book generated by ChatGPT prompts I wouldn't say it's essential to the points I'm making but I would say I would agree I think there are ways to use ChatGPT to do so. Somewhere between "the next letter is 't'..." and "write me a book" it goes from you creating a book to ChatGPT creating a book. I'd say the same for e.g. making a game, using stock engine templates doesn't mean you commissioned a game and you are no longer a game developer it means you can't claim that piece. At some point, just using everything out of the box leaves you with having created little to nothing though.

Using spelling and grammar checks is a great example, though maybe not as framed. I.e. it's not that "because professional writes use spelling and grammar checks" using any tool in any way is creating all the same it's an example of the idea when some part of the output didn't come from a direct, strictly controlled and intentional thought of the person it doesn't imply they no longer create the thing. I'd extend this to ChatGPT and say just because you have it spit out some words on how to word the next sentence doesn't mean the next sentence was no longer created by you at all. At some point though, the infamous "write me a book" style example, it must obviously stop being something you can claim as yours though. Like the art discussion, it's not a question of whether ChatGPT can be used by you to create a book (pop it in, ask for a spellcheck) or whether ChatGPT can write a book ("write me a book") rather at which point of relying on the tool makes the output creation more claimable by ChatGPT than by you.