AI/LLMs AI art bans are going to ruin small 3rd party creators

[kigmatzomat]

No one else would get to where I would get. The odds of hundreds of exacting details that 100% match my vision happening in one random go are longer than winning the lottery.

You avoid the point. If you had a clear vision and the competence to describe it the first time, you could get what you want with a single entry to an image generator.

A scenario you described as

I can enter a prompt and just let an AI run wild, having no control.

But you state the act of iterating is the delineator. What you call control.

I can ..... spend hours and hours, changing hundreds or even thousands of details via the AI tool, and have total control.

That's not control, that's iterating the vision and/or the verbiage. If you didn't know what color eyes you wanted on an elf with a bow, that's an incomplete vision. If you knew but didn't have the correct words to describe it, that is an issue with the verbiage.

You cannot make a lack of skill the factor that makes an action "creative". The ultimate AI-whisperer will enter "a" prompt and get what they need out of it.

What you are doing is learning how to formulate a complete concept and write AI prompts.

 

[Maxperson]

You avoid the point. If you had a clear vision and the competence to describe it the first time, you could get what you want with a single entry to an image generator.

100% false. I'd have to write a novel and even then it wouldn't be perfect. Things would be off just a bit and I'd have to tweak it.

That's not control, that's iterating the vision and/or the verbiage. If you didn't know what color eyes you wanted on an elf with a bow, that's an incomplete vision. If you knew but didn't have the correct words to describe it, that is an issue with the verbiage.

It's not an incomplete vision. The AI and/or myself can't know exact shades or shapes, so I'd have to tweak things. Not every shade has a name to be called, so lighter or darker would be a necessity.

You cannot make a lack of skill the factor that makes an action "creative". The ultimate AI-whisperer will enter "a" prompt and get what they need out of it.

And yet it can't match what I end up with. It's simply not possible to get the hundreds, if not thousands of details exactly correct with just a single prompt. We could both end up with elves that had bows, but the "whisperer's" wouldn't the same as mine.

 

[Lanefan]

If the work contains AI-made parts and human-made parts, then the AI-made parts are, well, made by the AI, hence derivative, and the other is made by the human and potentially creative. I don't see any merit in this reasoning.

Now if the AI-made part is now inexistent, it begs the question: in what precise way does the AI was useful in the process? What are we talking about, here? Please be as detailed as possible, because discussing all this in the abstract leads to confusion.

An example of something where the AI part would, in the end product, no longer exist:

Step 1 - get AI to produce a song or piece of music with each instrument/vocal on its own track
Step 2 - track by track, using the AI-generated music as a guide, play/sing a parallel track that sounds similar
Step 3 - delete the original AI-generated tracks
Step 4 - mix down only the human-generated parallel tracks to produce the finished song

The AI is useful in this process in that it provided the base song from which to work, and as that base song is not copyrighted it's free to plagiarize or copy like this.

To consider that the AI is simply a tool akin to a ruler that you just use as a guide for your hand, for instance, is simply untrue. AI doesn't work like that. It's not helping you making the thing, it makes the thing, and you don't even know how or with what. You just see the end result. So, unless you directing the AI for every pixels there is (at which point, frankly, just make the dam thing yourself), there will still be unaccounted-for pixels, pixels that was produced and placed there by the AI without you willing it. Pixels you didn't create.

In the above example, the AI is no more than a guide for the musician.

 

[Lanefan]

I'll try again. There is two possibilities:

You now how to draw a picture, in other words you know at least somewhat how you can materialize on screen an image you have in your head. In this case, you will be able to direct an AI to make this picture, which will be yours, because the AI was useless in this case. You could have made it yourself, with less efforts, at least in parts (and the parts you asked the AI to draw will not be your work).

OR

You don't know how to draw a picture. In this case you will NOT be able to ask the AI exactly what it needs to realize this vision of yours, because this vision doesn't yet have lines, textures, matter, techniques, framing, composition, nothing. All this will be provided by the AI and you'll just say yes or no, and the final work will NOT be yours.

You miss the third - and to this discussion most germaine - possibility:

You know how to imagine an image, but have no idea how to materialize on screen (or on any other surface) that image you have in your head. In this case, you will be able to direct an AI to make this picture because the image in your head is detailed enough that you can, with much experimenting along the way, eventually tell the AI exactly what to do*. You could not have otherwise made it yourself.

* - this assumes the AI is capable of understanding and interpreting highly detailed instructions, which I gather most are not - yet.

 

[BenjaminPey]

In the above example, the AI is no more than a guide for the musician.

And is also the original composer of the song, hence the song writing is simply derivative and not your work. You are only its performer.

You know how to imagine an image, but have no idea how to materialize on screen (or on any other surface) that image you have in your head. In this case, you will be able to direct an AI to make this picture because the image in your head is detailed enough that you can, with much experimenting along the way, eventually tell the AI exactly what to do*. You could not have otherwise made it yourself.

I strongly object to the possibility of such a thing existing. As I was saying, in order to give the AI instructions precise enough to obtain exactly what you have in mind (and all of this is built on the hypothesis that such thing is even possible, to have that kind of precise image in mind, which I would consider a very, very strong hypothesis) without any element previously introduced by the AI, you would need to have incredible drawing skills in the first place, just to express the instructions.

For instance and to keep it simple, if this image in your mind is like an oil painting, with bits in acrylic and other spray painted, you will need to know that, to identify that, and to express it in a manner so accute and precise that would be impossible to even approach or approximate without having mastered the skills in the first place. Hence, you already know how to do it yourself.

 

[BenjaminPey]

Case in point: let's take an image and let's say you have it "in mind".

This one, for instance (it's the last picture I saw on the Internet, not a cherry-picked one.)

What do you tell a very powerful AI which would perfectly understand natural language to make it produce exactly this picture?
(It's from Jesperish, if you're interested. They're very talented.)

 

[Morrus]

The AI is useful in this process in that it provided the base song from which to work, and as that base song is not copyrighted it's free to plagiarize or copy like this.

That's an interesting attempt at an end-run around copyright law so you don't have to credit or pay the original creators, but it doesn't work, I'm afraid. The AI work is not copyrightable, but it can still be a copyright violation. And there's plenty of lawsuits about that exact thing (from authors, and from the likes of Disney, and AI rip-offs of Darth Vader etc.) I'm not your lawyer, but I do not recommend this particular legal strategy.