These posts are getting huge, so let's define a few key theses that are going to be relevant throughout:
1. Very clearly someone directing a tool can be an artist. This tool can be advanced and substitute parts of the labor that may have otherwise been required. Example 1: a digital painter using Photoshop to execute their vision. Photoshop has plenty of features to help along, including AI algorithms, such as generative fill and smart masking. Example 2: a photographer using an iPhone. iPhone does plenty of things to improve picture quality, oftentimes WITHOUT any decision making from the photographer. For example, it will detect the presence of the sun and adjust brightness and colors accordingly. If you know anything about DSLR photography, you will know that it's a pretty massive part of taking a picture. There are loads more, these are just simple to understand examples.
2. Very clearly someone directing a person can be an artist. See my previous example of a movie director. In this instance, that person is NOT necessarily an artist themselves. For example, a camera operator can just be told how to execute the shot, without their input.
Also, since English language is complicated and the term "artist" has several meanings, I'm going to specifically refer to a person drawing something as a "painter" and a person who is performing some kind of creative endeavor as an "artist", to avoid confusion.
You're not reading what I'm writing. Even if the guy paying someone else to do it isn't the artist, the person performing or paid to do the art is.
See thesis 2. Either or both could be considered an artist. It depends on the situation. It could be that commissioner is very strict on their vision and the painter just does the work of putting it on the canvas, then the commissioner is the artist (the one doing creative work). It could be that the painter has complete creative freedom with just a general idea given, in which case the painter would be the artist. It could be that both contribute, making them both artists.
The algorithm reproduces patterns. It doesn't know what direction a human hand is supposed to face or how many fingers it's supposed to have or where the webbing of the hand is supposed to be. It just reproduces whatever shapes are commonly associated with a "hand".
I recommend you read a book on basic fundamentals of drawing. One of the first things you are taught is to never think about drawing "an object" and its relations in 3D space but rather think about simple 2D shapes that you see. Painters quite literally disable their depth perception on purpose by closing one eye when observing their subjects. The famous training exercise of drawing upside down is meant to force you to stop thinking about real world concepts and focus on the shapes. There are some techniques that have to do with thinking of simple 3D shapes (like spheres, cylinders, cubes) on the canvas and constructing the subject from there, but those are in no way a requirement to paint.
We could also talk about how neural networks do actually quite literally get an understanding of real world concepts - an easy to understand example would be that embeddings of language models words such as "king" and "queen" will be in the roughly in the same relation to each other as words "man" and "woman", despite being in different parts of the latent space. Meaning the model will understand the concept of gender. But it's not super important to this particular point. It's just that the model will get whatever information or concepts necessary to perform its task by nature of training, sometimes better than humans (this is how we lost to bots in chess for example, or discover new mathematical proofs via AI).
Hands by the way are famously difficult to draw in perspective, especially in arbitrary poses. Remember I talked about people mistakenly witch hunting actual painters? Well that often comes from hands not looking perfectly acceptable. And not only that, but AI will of course be affected by any errors in the dataset that was used to train it. In that sense, AI was bad at hands because humans are bad at hands! Isn't that cool? Notice I said "was", that's because newer models have been pretty good at hands for quite a while now. Finally, if you see extremely bad hands on a gen, that's because the author didn't put any effort into fixing them up. With inpainting it's extremely easy to do. This will be relevant later.
Regardless of how much effort you put into something or like doing something, there's still conscious thoughts behind those decisions. If you're commissioning a piece of art, you can ask the artist why they've chosen what they've chosen. They have an opinion of what they've made, or why they decided to use the style that they use. The art is undoubtedly the work of the person who produced the work. And to clarify, I never said that human work is inherently creative, good and devoid of bad qualities. I'm expressing that AI doesn't make or decide anything on its own whim, or in other words, it's soulless. The person having AI make something for them still doesn't actually do any of the work themselves. The words or images that appear on the page aren't something that came out of their head. Just because they ask the machine to revise it until it gets closer to their vision still doesn't change that. I don't know the first thing about AI being used in photoshop and I don't intend to pretend to.
I mean you can study why a model made a particular "choice", we do all the time. It's just difficult, it's like if the painter was in another language. Speaking of languages, humans also do things rather "mindlessly" oftentimes. For example, native speakers often struggle to explain rules of their language to a learner, saying something like "I dunno man, it's just how to is". That's because they internalized it through practice, rather than consciously thinking about grammar rules. This is what becoming "fluent" essentially is. This applies to many skills where they kind of become automatic rather than being conscious. Painting is no exception. One of the best times I was painting was where I was in the "flow", rather than thinking about it consciously. And to be clear, no one is claiming that AI is conscious. I consider it a tool, but a clever one at that, and one that has many attributes of a person executing your decisions for you. Makes sense, since it does have neural networks at its heart.
That would depend on what you mean by design. Do you mean by recipe, physical appearance, or what? But regardless of who is the creative party, both of you have an idea of what looks and tastes good. A chef could intentionally come up with a flavor that is pleasing even when it isn't conventional, or improvise something to replace a missing ingredient, based on their knowledge and skill. If AI could understand or learn to understand why their art oftentimes looks so uncanny, their art wouldn't look so uncanny. It can't even initiate its own research into the topic.
I don't know why this is important. The point is basically thesis 1. Not sure why you'd want a tool to do research, or what you mean by that.
I don't think you and I are thinking about the same thing. I'm talking about the delusional people who think they have a novel idea and expects someone else to happily do all the work to make it a reality because the idea is inherently valuable, but doesn't intend to pay them until the idea succeeds.
Then I'm not sure what your example has to do with the discussion, or what it was supposed to illustrate.
I can only agree to disagree. If we get into this, we're just going to talk in lengthy circles about semantics and further clutter the thread.
As long as we agree that it has nothing to do with copying or tracing or editing or theft, then I'm good with this. Just don't want to pollute the discourse with inaccurate emotional language.
I'll be the first to admit that I'm not exactly a great orator or expert in AI, but if we're talking about human perceptions, there's definitely something about AI, even good looking AI, that feels off. It's the smell, if there's such a thing. But just because some people unfortunately draw like AI doesn't mean that there isn't something uncomfortable about AI art. That said, as an amateur creative type, there's definitely things you pick up on and appreciate that I didn't before I tried my hand at creating, and it's not the same as browsing for something that suits my taste.
You cannot possibly how well you detect AI art because, well, you cannot possibly know anything about the ones you don't detect. It's been very possible to generate perfectly good artwork that vast vast majority of people won't be able to detect for a while now. It just takes more effort and knowledge. The ones you see as "obvious" are low effort gens. Just like regular artwork, bad gens are more prevalent. I have personally seen a couple of people pass themselves off as normal artists to take on commissions. Their works are posted in public on websites that banned AI gens. Believe me, I'm no less upset by low effort gens than you are. I'm equally as upset by low effort everything though. I mean again, we are on f95.
