nice line though hehe 
Typing in a prompt and picking from auto-generated results is, in my opinion, just one step above going to an art store and looking for a picture that would look good on your wall. To refer to such a thing as true artistry is as much a stretch as those thumb twitching Starcraft players demanding they be called athletes (because of the emergence of e-sports).
1 Like
Sure, typing in a prompt and picking the first decent-looking image isn’t artistry in any sense of the word in my opinion either. I prompted some Disco Diffusion images that I quite like; they’re pleasing to the eye, and that alone might qualify them as art for people who don’t know their provenance. But I don’t feel I expended much effort or conscious guidance of the process, so no, this isn’t artistry and doesn’t make me an artist in my own assessment. (I generally don’t apply the term to myself in non-AI creative domains for … reasons.)
But I think the line gets quite blurry when somebody spends weeks carefully and purposefully crafting that prompt and fine-tuning the images according to whatever visions dance around in their head. It seems possibly just another method to get what’s in your head into a more tangible form. That’s not the same thing I did at all, and I won’t dismiss it as easily. I hope during this debate we’ll have lots of conversations with people who do consider themselves artists but lean heavily on computer-generation in their process, whether it be AI or those methods that came before.
Was Ansel Adams an artist? He didn’t create any of his subject matter.
1 Like
it’s already possible, there are already people making use of these diffuse models to create textures of all kinds
I am unconcerned that computers will ever have “creativity” or “imagination.” Nor do we need them to. For the most part, “AI” techniques try to reduce the amount of “manual specification” that you must do to describe the outcome that you want. Early techniques were “100% manual” and today most of them still are. Gradually, we’re getting things like “CryptoMatte” and “crowd simulation,” but there’s still a long way to go.
The way I see it, “doing stuff so that I don’t have to” is what a digital computer is for. I’d love to be able to just “describe” what I want, let the computer churn away for a while, and show me something so I can decide if I like it. If that “churning” just did “what used to be a week’s worth of work for me,” then I am a Happy Boy.™
The only thing that I still don’t like about CG is how labor-intensive it presently is … and how much of that labor just feels to me like drudgery. Give me power tools. Let the computer itself be stuck with the drudgery.
4 Likes
I am not sure whether I agree with that. It certainly depends on how you define those terms. Hypothetically, let’s assume we have something like Stable Diffusion which allows you to generate lots of variants, based on the seed, but also using similar texts. Besides that, there would be a classifier to find images of our (the user’s) liking and some sort of a clustering algorithm(s) to find the more unique or outstanding images. We might let it experiment on its own to find unique examples, which it could then further refine to make the more likable.
I am convinced this sort of system has the potential to come up with solutions which many humans would describe as creative.
Quite a few would argue the current system comes up with solutions that are creative. Arguments that it’s not creative often rely on the fact that it’s a merely synthesis of other peoples ideas and not wholly unique (an argument which leaves most humans incapable of creativity if we’re being honest with ourselves) or rely on the fact the computer doesn’t have the self-awareness to create with intent.
If we accept that as a necessity for creativity, we have a system that can produce works perceived as creative yet can’t be labeled creative due to a technicality, one that may be dependent on the perceiver’s ability to recognize it as AI work. Many can, but some don’t recognize the signs.
On the other hand there’s plenty of things made by human creators with self-aware creative intent that nonetheless aren’t particularly creative by other people’s standards. Probably a philosophical issue more than anything else.
3 Likes
That’s true. In fact, what I described also relies on this. It would in a sense filter for more unique examples and try to make them meet some quality standards.
I think you are the first person who directly explains to me (at least one reason) why self-awareness might be relevant. Thanks a lot for that! I need to think more about it. (But it sort of feels dishonest at first. Like moving the goal post towards more and more abstract and philosophical reasons to preserve it.)
The idea that if a computer and a human came up with the exact same image and one would be considered creative (or art) and the other one not still baffles me. To me, it is like not even allowing a dialog to happen, because by some old definitions, computers can’t be creative or create art.
In my opinion it is not a philosophical issue. It is highly individual and as such there is no clear cut as to what is creative or what is art. As a consequence, trying to cover creativity or art in a definition is very difficult and likely the easiest way to achieve it is by using abstract descriptions and some special pleading for humans (to make us feel better about ourselves).
Thinking about this again, to me, it is more like: Looking at it from a philosophical point of view makes this subject unnecessarily complicated and abstract to talk about it.
1 Like
I honestly lean more towards viewing generative art as creative art for pieces I would use those words for when made by humans because it seems a bit reductive to insist generative art can never be art because it’s generative. But I can see both sides of the argument.
That’s partly why I say it’s largely philosophical actually: philosophical issues are highly individual and there is no clear cut answer to them.
3 Likes
I wasn’t very clear regarding my description apparently. Contrary to most philosophical topics, when it comes to creativity or art, they are to a certain degree testable. Even though one can look at them from a philosophical point of view, I simply don’t see a lot of value in it.
From my point of view, it would be more honest to consider something as creative or art, if people perceive it as that. And this is easily testable.
In reading thru these topics there’s a point it looks like nobody else has made (sorry if I’ve missed it). Maybe it’s come to me because I’m both an artist and a fiction writer (both mostly amateur). It seems to me that most visual-media artists, in a “when you’re holding a hammer every problem looks like a nail” sense, don’t see an evocative verbal description as art.
From Wikipedia: “ ‘art’ is commonly understood to be skill used to produce an aesthetic result . . . skill is being used to express the artist’s creativity, or to engage the audience’s aesthetic sensibilities.” When thinking about pre-computer sculpting and painting, it’s easy to think that physical skill is a necessary component of art. The years of practice learning the tools of the media, be they chisel, brush, or mouse – if that’s not art itself, it’s a fundamental component. Art must flow thru the hands (or close equivalent) to truly be art.
But writers’ understanding of artistic skill is different. Our art is, essentially, expressed thru our breath – the art of the storyteller. The tools facilitating that thru the ages, stylus’n’clay, ink’n’paper, keyboard’n’pixels – and now “AI generated” from a verbal description. The AI is a person’s tool, much like all the other tools, easing our ability to make art, but for all its near-magical output, it’s still a tool. One which requires specific skills – the writer’s skill of description being just one of them. I could also cite the photographer’s (yes, photography is an art form) ability to select composition, lighting, etc from the surrounding environment in the choices made of which of the AI’s outputs fulfill the artist’s intent.
If a paint splash is art, that requires the artist’s intent – a bucket fallen from a ladder can’t make art, it takes a person saying “yes, this”. The problem of telling the difference between an artistic paint splash and a mundane accident has been with us for decades, the latest AI stuff is just the latest version of it.
Although I’d agree that the perception of something as art is a fundamental component, I’ve not seen that as “easily testable”. Clarify pls?
1 Like
I am still skeptical about the intent idea being humans. It is fairly easy to build a system that explores on its own, guided by metrics such as uniqueness or appeal. Those might even be changing, and as such moving away from the original intent to its own one.
Show a person an image. Now that person can answer the question whether they find it creative, artistic, art or whatever.
Which is why I said “person”, not “human” (although I’ll grant you others have). I’d say that art can be made by / appreciated by any person – if a baboon with a camera, an octopus with a cellphone, or an AI is sufficiently sapient/sentient to be a person, then there’s artistic potential. No people involved? Therefor not art – failed to meet one of the minimum criteria.
Oh. Gotta disagree there. Significant consideration in experimental design needs to be made to compensate for perception ability, cultural differences (including, but not limited to, bad faith responses to meet cultural expectations), bad faith responses to meet personal expectations (“Am I supposed to think this is art? Don’t wanna look uncultured, better say yes!”), etc. Given the currently known issues with designing such a study? “Testable” I’ll go with. “Easily testable”? Not even remotely, sorry.
1 Like
That’s the minimum criteria which needs to be rethought in my opinion. If I am presenting an image and I claim that I made it, then it is art. If I say it was generated by a computer, now it isn’t anymore.
Why is being sapient/sentient a requirement in your opinion?
I meant easily testable as in: The methodology how to properly test it is known. It doesn’t need to be invented first.
There’s a lot of discussion across multiple topics, so sorry if I missed you saying earlier: what definition of of “art” are you using? 'Cause I’m going with what I quoted above:
From Wikipedia: “ ‘art’ is commonly understood to be skill used to produce an aesthetic result . . . skill is being used to express the artist’s creativity, or to engage the audience’s aesthetic sensibilities.”
You can see from this that art requires at least one person to make the art and at least one person who perceives it as art (since both can be the same person, therefor at least one person). I don’t require this person to be human, but I do require sufficient sapience/sentience for personhood.
But that’s me. Not you? What’s your definition of art, what’s your interpretation of that definition that allows for nonsapient/nonsentient machines to make it? (And if you’ve already said, sorry for not keeping up, I’d appreciate a link).
1 Like
To me, the human who perceives it is the judge whether they consider it art or not. Going way beyond that makes the subject unnecessarily complicated in my opinion.
That’d be fine for “is it pretty?” or even “is it beautiful?”, but is overly simplistic** for “is it art?” by the commonly understood meaning of the term.
** I was gonna put “IMHO” here, but given the over-a-century’s worth of discourse on the subject that Wikipedia cites, looks like it’s not just my opinion, it’s a general consensus, despite plenty of other extant disputes.
It is consciously oversimplistic. I admit I went a bit too far with it 
I also agree with this sort of definition (from the wikipedia article):
By its original and broadest definition, art (from the Latin ars , meaning “skill” or “craft”) is the product or process of the effective application of a body of knowledge, most often using a set of skills;
I struggle to agree with people who add consciousness or other characteristics that are usually mostly common in humans. To me, what is important is the end result. My assumption is that people mostly associate art with certain skills and only appreciate something as art, if they can see that skill set being applied.
1 Like