AI Created images (generated by Midjourney) lose registered copyrights

[jagdtigger]

46 minutes ago, Holmes108 said:

Humans are trained on copyrighted works and can ultimately make art that is sufficiently unique.

Taking inspiration and training are wildly different things......

[jagdtigger]

37 minutes ago, Kisai said:

That's describing how it works with a very naïve understanding of it. The "original" data is not in the model.

 

Yeah right... /s Thats why watermarks and other characteristics surface very easily. AI is just a big@ss lookup table of pre-made things which get stitched together by algorithms without any sort of creativity. Meaning there is nothing that warrants any sort of protection for the generated image/text/etc.

[Kisai]

20 minutes ago, jagdtigger said:

 

Yeah right... /s Thats why watermarks and other characteristics surface very easily. AI is just a big@ss lookup table of pre-made things which get stitched together by algorithms without any sort of creativity. Meaning there is nothing that warrants any sort of protection for the generated image/text/etc.

 

Again, that's not understanding how it works, and they don't surface easily except when overfit. SO it you have 1000 images from GETTY News in the training data, and they all say "GETTY NEWS" in the text, then of course that's going to surface from the overfitting. If you type "dog" you're not going to see watermarks on dogs, only sometimes with "GETTY DOG."

 

Here's another example of not understanding how it works:

 

Again, completely does not understand how this works, and the user who made the image is doing the exact same thing they accuse the AI of doing. 

 

Everything you input into an inference mode of a model is immediately lost when you leave. There is no "also training the AI". No that only happens if you are uploading images to the AI platform without reading the terms of service that they will do that. Are they? I don't know, maybe read it yourself. But they sure as hell aren't being used to train the existing model you are using.

 

The AI is not going to learn anything from submitted data any more than it will learn pronouns or propaganda. The only reason any of this stuff shows up in inference is because it has learned the word association with the weights. Nothing more. There is no meaning behind it.

 

[LAwLz]

4 hours ago, jaslion said:

Gets the money out of them. Which means it's a less desirable now to overtake this space. Which means the space stays open for humans to enjoy without big corporations pushing everyone out.

Gets the money out of who exactly?

Why does not being able to copyright an image makes it less desire to "overtake this space"?

 

 

I would argue that this makes it more ripe for big corporations to push out artists, and makes the space less open for humans to enjoy.

I think a lot of artists has frame this whole discussion as "us, the human artists, vs them, the corporations using AI", but the actual battle is in my eyes against people with a valuable skill (making art) trying to step on the average Joe who do not have the skills but instead have to rely on tools to assist them in creation.

 

The person this entire lawsuit revolved around was a single woman who wanted to make a comic book. She is the loser here. A writer who had an idea for a comic book but might not have had the financial opportunity to hire artist(s) to work on the comic book. With the help of AI tools she was able to create the book anyway, for cheap, by doing the work herself.

 

 

I think a lot of artists are trying to deceive the general public by framing this as "AI is used by massive corporations trying to eradicate artists!".

From what I can tell, most artists who are against this are those who do a lot of commissions for the average Joe. People like those who get commissioned to draw a D&D character someone requested, or someone who gets commissioned to design some logo or something. Those seem to be the majority of people who complain, and those people are just complaining because the average Joe, the actual underdog, are suddenly able to do things for themselves, without having to rely on artists.

 

 

 

I genuinely believe that a lot of artists are against this because they want to gatekeep a form of expression. They don't want "commoners" to be able to do what they spent years teaching themselves to do. They want to keep people down so they themselves can be above others in some way.

 

 

 

3 hours ago, Kisai said:

Asking the wrong question. That should be "what is the minimum threshold of human input required?

That's what I meant. My question was suppose to be rhetorical.

My question was more along the lines of "what do you think should be the minimum threshold".

I already know what the court has ruled  and I disagree with it. I think, as the lawyer said, that this ruling happened because the people who judged it doesn't really understand the technology. I think a lot of people who comment on this doesn't understand the technology and the amount of human input and work that is needed to get what you want out of it.

[LAwLz]

3 hours ago, goodtofufriday said:

I think your arguement generally fails because its humans interacting with humans with humans doing all the interpreating. 

If we had the food making machines from star trek, would you call someone who told a machine "make a michelin 5 star meal" a chef? Probably not. So why is someone who tells a machine "make this piece of art" an artist? 

It all about humans hands or mind physically interacting with the space around us, taking bits and pecies of things to create something new. A machine doing the bits and pieces part is not the human creating. 

So a human interacting with a machine who does interpreting is not art?

Does that apply to cameras, CGI, Photoshop and other programs too or just anything that has the word "AI" in it?

 

 

3 hours ago, goodtofufriday said:

It all about humans hands or mind physically interacting with the space around us, taking bits and pecies of things to create something new. A machine doing the bits and pieces part is not the human creating. 

Is inputting commands into a computer not "human hands or mind physically interacting with the space around us"?

If the answer is no, what difference is AI art generation vs using functions in photoshop to edit an image, or clicking a shutter button on a camera?

[LAwLz]

3 hours ago, Holmes108 said:

Not sure if this has been brought up yet, but what about the danger of AI becoming the equivalent of a copyright/patent troll. What if I can make a computer spit out literally millions (maybe billions?) of pictures in the span of a day. Even if they pass the "originality" test for us all... they could perhaps come too close to owning all too many variations of expression, breaking the system...

It has already been brought up, and it has been attempted even without any involvement of AI.

 

Someone brought it up here:  

5 hours ago, BabaGanuche said:

The problem with having AI getting copyright is problems like this, https://www.vice.com/en/article/wxepzw/musicians-algorithmically-generate-every-possible-melody-release-them-to-public-domain, where the AI exhaustively copyrights a particular area.

 

The TL;DR is this:

All music that exists and can ever exist is comprised of only 8 notes, just rearranged in different orders.

Someone wrote a program that through sheer brute force, generated every single combination of those 8 notes, thus creating every single melody that can ever exist.

I am fairly sure it didn't go anywhere because it's so absurd.

 

 

I think copyright law needs to be updated to account for AI tools. It has been utterly broken and favoring big corporations for way too long so I'd argue that it needed to change even before AI tools, but it definitely needs to get updated now. It would be very easy to just add clauses that makes "copyright trolling" by generating thousands of different iterations not valid.

 

New technology almost always result in rules, regulations and laws needing to adapt. We shouldn't get too caught up in "the system we have today wouldn't work if X technology existed so let's kill technology X". What we should focus on is "how can we adapt our system to account for technology X?".

[LAwLz]

3 hours ago, wanderingfool2 said:

It depends I think.  The issue about having "AI art" non copyrightable is that it can get very murky in terms of where the line is drawn and it can take away some of the rights people might expect.  Specifically around how much user interaction should be required for something to be copyrightable.

 

As an example, if they trained it only on images they drew should it be copyrightable? (I'd argue yes)

Now what happens if lets say you provide a skeleton frame that you wish to animate and you use AI and a reference image to do it?

What if you slowly describe and inpaint an image to get an image you want?

What if you describe an entire scene and it makes that scene, is it specific enough?

 

The issue I have with AI art being non-copyrightable is that I do believe there are some things that should be copyrightable (maybe they would lose certain aspects of copyright, but I think it shouldn't be outright not copyrightable). 

That's exactly the answer I was thinking of when I asked that question.

I think putting a "this is not copyrightable" blanket on any and all images generated through a program like stable diffusion is a very bad thing. The line should be drawn somewhere, and right now it seems like the line is drawn at "if the program mentions AI, it's not copyrightable" because the people who make these rules don't understand how it works.

 

I think we need a more nuanced approach, but putting up hard criteria to define where the line is draw will probably be very hard. That's why I am so against people who are just saying "this is good" because I don't think they understand the varying degrees of effort and human input that can go into "AI art".

AI art is not a single button program that you just click and get exactly what you want out of, just like Photoshop isn't a single button program that just magically makes any and all pictures better.

[goodtofufriday]

37 minutes ago, LAwLz said:

So a human interacting with a machine who does interpreting is not art?

Does that apply to cameras, CGI, Photoshop and other programs too or just anything that has the word "AI" in it?

 

 

Is inputting commands into a computer not "human hands or mind physically interacting with the space around us"?

If the answer is no, what difference is AI art generation vs using functions in photoshop to edit an image, or clicking a shutter button on a camera?

Inputting commands it not the same no. Using a digital pen, or mouse pointer, acts in the same way a real pen does. Modeling a model in blender is substancially similar to taking real clay and making a model. Cameras capture moments that are either candidly made by interactions of living things or scene that were created by living things. You cant click a shutter button and have a picasso level photo. 

But AI "art" removes all of the above. It does not behave in a similar maner to any physical tool that humans attribute to creating art. These commands are nothing more than an idea. Ideas are not subject to copyright. 

If i have an idea for a photo i need to physically make or find the scene in order to reproduce whats in my minds eye. 

 

Your arguments for these other tools all involve humans needing to actively do something beyond inputting commands that amount to just ideas.

Its a disengenious argument at best. 

[LAwLz]

3 hours ago, jagdtigger said:

Because its a generator and not a creator. It takes ppls work and use parts from them without any sort of modification and stitches them together......

I don't generate an image when I take a picture with my camera. My camera does the generation for me. I give some vague inputs and the camera does the work of actually inferring, computing and outputting the image. In the case of for example landscape photograph I can't even claim to have constructed the subject. I am just telling a tool to capture what nature created.

And yet, nobody is questioning the fact that I get attributed copyright to those images.

 

 

2 hours ago, Brooksie359 said:

One thing I could think of is that the AI potentially could spit out the same image more than once and then who would own the copyright? I mean technically the copyright should go to the AI itself and not someone who is simply utilizing the AI. 

This is already an "issue" with any and all software.

Digital audio programs can spit out the exact same music as someone else have created unintentionally. The likelihood of the same piece of work being generated is next to none.

Even if someone were to use the exact same model, using the exact same prompts, with the exact same settings, you'd still need to happen to hit the exact same seed as well.

Looking at Stable Diffusion's source code, we can see that in the backend, the seed is an unsigned long. So in Stable Diffusion, there is an artificial limit of 18446744073709551615 different images that can be generated using the exact same model, prompts and settings.

 

So it's 1/18446744073709551615 that the same image gets generated (less than one in 18 quintillion). That is probably less risk than the same piece of music being generated.

 

 

 

3 hours ago, jagdtigger said:

As for the topic its a good decision. Now the only thing we need is a ban on using copyrighted material without permission as training material.

I've seen this get suggested several times and it is one of the stupidest things we can do.

If we banned using copyrighted material for training without explicit permission do you know what would happen? Only companies like Disney, Facebook, Google, etc could train their models, because those are the only ones big enough to create datasets fromt heir own data they have the rights to.

All of a sudden you have made sure that no regular person can use AI to generate art. You have made it so that only big companies can do it. Which is exactly the opposite of what most artists want.

 

It is very important that everyone have the ability to train AI models because the alternative is that only large companies can do so, and we certainly don't want to give massive corporations even more benefits when it comes to competing with small independent creators.

[Brooksie359]

6 minutes ago, LAwLz said:

I don't generate an image when I take a picture with my camera. My camera does the generation for me. I give some vague inputs and the camera does the work of actually inferring, computing and outputting the image. In the case of for example landscape photograph I can't even claim to have constructed the subject. I am just telling a tool to capture what nature created.

And yet, nobody is questioning the fact that I get attributed copyright to those images.

 

 

This is already an "issue" with any and all software.

Digital audio programs can spit out the exact same music as someone else have created unintentionally. The likelihood of the same piece of work being generated is next to none.

Even if someone were to use the exact same model, using the exact same prompts, with the exact same settings, you'd still need to happen to hit the exact same seed as well.

Looking at Stable Diffusion's source code, we can see that in the backend, the seed is an unsigned long. So in Stable Diffusion, there is an artificial limit of 18446744073709551615 different images that can be generated using the exact same model, prompts and settings.

 

So it's 1/18446744073709551615 that the same image gets generated (less than one in 18 quintillion). That is probably less risk than the same piece of music being generated.

 

 

 

I've seen this get suggested several times and it is one of the stupidest things we can do.

If we banned using copyrighted material for training without explicit permission do you know what would happen? Only companies like Disney, Facebook, Google, etc could train their models, because those are the only ones big enough to create datasets fromt heir own data they have the rights to.

All of a sudden you have made sure that no regular person can use AI to generate art. You have made it so that only big companies can do it. Which is exactly the opposite of what most artists want.

 

It is very important that everyone have the ability to train AI models because the alternative is that only large companies can do so, and we certainly don't want to give massive corporations even more benefits when it comes to competing with small independent creators.

I guess I just find it weird that one could tell an AI what to do and it does the work and you would get the copyright. I would imagine that the one who created the AI should be the owner of the copyright tbh. But then again when someone trains an AI using open source code then it really is a Grey area. Anyways I am not a super strong proponent of copyright in the first place so I do not see why it matters anyways. 

[divito]

AI art should be copyrightable. I think some people's logic is pretty fallacious in denying copyright. Photography, or editing with things like Lightroom are practically hands-off. Change 0-5 settings on your camera, take the picture, click a filter in Lightroom and BAM, you're super skilled and can copyright a photo.

If it helps assuage people, perhaps the code/engine/model can be included in the copyright alongside the individual that used the prompts and other inputs and choices that developed the image, no different than a photographer, or someone that uses filters and tools in Photoshop or Lightroom.

[Arika]

 

6 minutes ago, goodtofufriday said:

Your arguments for these other tools all involve humans needing to actively do something beyond inputting commands that amount to just ideas.

Its a disengenious argument at best. 

since you boil it down to something so simple. i want you to tell me what prompts (and negative prompts) you would use to essentially recreate this picture assuming same model, seed and hypernetwork since apparently it's just inputting commands with no effort and we'll see how well your idea translate to reality

 

Spoiler

 

this base prompt i have been tweaking over the course of 2 months and thousands of images to get it to where it is. and this isn't even one of the models i've trained myself.

[LAwLz]

Just now, goodtofufriday said:

Inputting commands it not the same no. Using a digital pen, or mouse pointer, acts in the same way a real pen does. Modeling a model in blender is substancially similar to taking real clay and making a model. Cameras capture moments that are either candidly made by interactions of living things or scene that were created by living things.

First of all, why when and why did "behaves like an object in the physical world" become a criteria for something being eligible for copyright? 

Also, fuck no, a digital pen or mouse pointer does not act the same way a pen does. There is a reason why Wacom made a special lineup of drawing tablets that works with a real pen. They made that because a stylus does not behave like a real pen.

 

And modeling in blender is in no way shape or form like modeling with real clay. You got to be fucking kidding me right now. Have you ever tried modeling with clay and have you tried blender? They are nothing alike.

Michelangelo was a famous sculptor. Are you seriously going to pretend like he would be good in Blender just because he knows how to create marble statues? That is ridiculous and you know it. Blender does not behave like clay sculpting any more than AI art generation behaves like a camera or painting. You are being disingenuous.

 

Also, a camera does not just capture things created by living things. That was even a point I made in my post. A picture of a scenery is still copyrightable even though it was not created by a person or other living thing. A picture of the Grand Canyon for example does not necessarily contain anything that was by interactions of living things.

 

Also, the entire argument breaks down when we start taking things like shortcuts into consideration. Even if we accept the ridiculous argument that a mouse pointer behaves like a pen, what about shortcuts? 

Taran who used to work for LTT hade so many shortcuts when he did video editing that he needed a secondary keyboard. What physical tool is pressing "H" on a keyboard to execute a script equivalent to?

 

 

Just now, goodtofufriday said:

You cant click a shutter button and have a picasso level photo. 

So? What does that have to do with anything? It's still eligible for copyright even if it's a shitty photo.

I feel like we are slowly moving into the whole "I don't consider this art because X, Y, and Z" which is not something I feel like discussing because it just becomes a bunch of feelings and is not related to the topic at hand.

 

 

Just now, goodtofufriday said:

But AI "art" removes all of the above. It does not behave in a similar maner to any physical tool that humans attribute to creating art. These commands are nothing more than an idea. Ideas are not subject to copyright. 

So what? Why does it matter that AI art "doesn't not behave in a similar maner to any physical tool"?

That's an entirely arbitrary restriction you came up with. Plenty of functions in Photoshop does not behave like real life tools either. Does that mean artwork created in Photoshop using those tools shouldn't be copyrightable or considered art?

 

 

Just now, goodtofufriday said:

If i have an idea for a photo i need to physically make or find the scene in order to reproduce whats in my minds eye. 

Or create it by using a computer program.

 

 

Just now, goodtofufriday said:

Your arguments for these other tools all involve humans needing to actively do something beyond inputting commands that amount to just ideas.

People need to actively do something to use AI art tools as well. Input commands, tweak settings, and so on. Just like someone using Blender needs to input commands, tweak settings, and so on.

[LAwLz]

Just now, Brooksie359 said:

I guess I just find it weird that one could tell an AI what to do and it does the work and you would get the copyright. I would imagine that the one who created the AI should be the owner of the copyright tbh. But then again when someone trains an AI using open source code then it really is a Grey area. Anyways I am not a super strong proponent of copyright in the first place so I do not see why it matters anyways. 

I mean, isn't that what we do all the time with computers?

When I apply an effect using After Effect, I tell a computer what I want it to do, and it does all the computation for me automatically in the background. I don't think it would make sense to attribute copyright of a video I edited in After Effect to Adobe just because they wrote the program which did the work for me.

[goodtofufriday]

12 minutes ago, Arika S said:

 

since you boil it down to something so simple. i want you to tell me what prompts (and negative prompts) you would use to essentially recreate this picture assuming same model, seed and hypernetwork since apparently it's just inputting commands with no effort and we'll see how well your idea translate to reality

 

  Reveal hidden contents

 

this base prompt i have been tweaking over the course of 2 months and thousands of images to get it to where it is. and this isn't even one of the models i've trained myself.

I'll whole heartedly admit I can't make a prompt that would create that most likely. 

But that's because the idea you had was uniquely yours. 

Laws have firmly established that ideas can't be given copyright. 

 

And how is this any different that someone telling a food making Machine to use less salt and more pepper, 1 hour cooks vs 1hour 10min.  They didn't cook the food themselves, and are there for not a chef. 

 

You give the machine your idea. The machine creates the idea, in this case an art piece. You yourself are not the artist.  But machines is. They can't be granted copyright. 

 

[Arika]

10 minutes ago, goodtofufriday said:

I'll whole heartedly admit I can't make a prompt that would create that most likely. 

But that's because the idea you had was uniquely yours. 

Laws have firmly established that ideas can't be given copyright. 

 

And how is this any different that someone telling a food making Machine to use less salt and more pepper, 1 hour cooks vs 1hour 10min.  They didn't cook the food themselves, and are there for not a chef. 

 

You give the machine your idea. The machine creates the idea, in this case an art piece. You yourself are not the artist.  But machines is. They can't be granted copyright. 

 

i'm not talking about the copyright side of thing. i'm referring to your point that creating AI art is just "inputting commands" and that there is no human actively involved.

 

Quote

But that's because the idea you had was uniquely yours. 

you don't need the idea, the image is right in front of you. how is an image in my head and figuring out a way to get an output any different that you seeing the finished product and figuring our how to recreate the output?

[goodtofufriday]

7 minutes ago, LAwLz said:

First of all, why when and why did "behaves like an object in the physical world" become a criteria for something being eligible for copyright? 

Also, fuck no, a digital pen or mouse pointer does not act the same way a pen does. There is a reason why Wacom made a special lineup of drawing tablets that works with a real pen. They made that because a stylus does not behave like a real pen.

 

And modeling in blender is in no way shape or form like modeling with real clay. You got to be fucking kidding me right now. Have you ever tried modeling with clay and have you tried blender? They are nothing alike.

Michelangelo was a famous sculptor. Are you seriously going to pretend like he would be good in Blender just because he knows how to create marble statues? That is ridiculous and you know it. Blender does not behave like clay sculpting any more than AI art generation behaves like a camera or painting. You are being disingenuous.

 

Also, a camera does not just capture things created by living things. That was even a point I made in my post. A picture of a scenery is still copyrightable even though it was not created by a person or other living thing. A picture of the Grand Canyon for example does not necessarily contain anything that was by interactions of living things.

 

Also, the entire argument breaks down when we start taking things like shortcuts into consideration. Even if we accept the ridiculous argument that a mouse pointer behaves like a pen, what about shortcuts? 

Taran who used to work for LTT hade so many shortcuts when he did video editing that he needed a secondary keyboard. What physical tool is pressing "H" on a keyboard to execute a script equivalent to?

 

 

So? What does that have to do with anything? It's still eligible for copyright even if it's a shitty photo.

I feel like we are slowly moving into the whole "I don't consider this art because X, Y, and Z" which is not something I feel like discussing because it just becomes a bunch of feelings and is not related to the topic at hand.

 

 

So what? Why does it matter that AI art "doesn't not behave in a similar maner to any physical tool"?

That's an entirely arbitrary restriction you came up with. Plenty of functions in Photoshop does not behave like real life tools either. Does that mean artwork created in Photoshop using those tools shouldn't be copyrightable or considered art?

 

 

Or create it by using a computer program.

 

 

People need to actively do something to use AI art tools as well. Input commands, tweak settings, and so on. Just like someone using Blender needs to input commands, tweak settings, and so on.

This is what I mean by disingenuous. 

You want to argue now how a pen and paper aren't an exact 1:1 to a digital pen. 

The point is both instruments draw a line. Draw enough types of lines and you get an image. 

 

You want to argue how blender and clay aren't 1:1. 

The point is both require you to be able to mold a form into something. The creative process of it all.

 

You either miss or entirely ignore the artistic and creative process. 

 

A photo of an existing canvas can't be copyright. Not all examples of photos can be made copyright. 

A photo of the grand canyon involves where the sun is hitting, what angle and elevation you're standing at, type of lens, zoom levels. Etc. 

All those elements create a scene that was made by a human. Intentional or not. And thats what makes a photo copyrightable. It's not because it's simply a photo.

 

And if I'm not mistaken didn't taren create a lot of those short cuts himself? Short cuts that are definable? A series process that human made? 

Doesn't sound like a ai generative process there to me.

 

And to your point for Photoshop, I would actually argue that some of those tools make an extremely thin line as to what's copyrightable and what's not. Like brushes that basically draw the hair for you? I don't think that should be considered human art. 

But that's my personal take on that and outside of this argument. 

 

Actively typing in an idea to a program that does the creative process for you does not rise to the level copyrightable work. Because said person is not doing the creative process work themselves. 

 

We obviously exist in two different realities here. Arguing with you further is only going to be an argument, not a constructive discussion. So I'm going to walk away here.

[goodtofufriday]

3 minutes ago, Arika S said:

i'm not talking about the copyright side of thing. i'm referring to your point that creating AI art is just "inputting commands" and that there is no human actively involved.

 

you don't need the idea, the image is right in front of you. how is an image in my head and figuring out a way to get an output any different that you seeing the finished product and figuring our how to recreate the output?

I made a example earlier if skilled work that was done to a car. That modified car was not granted a copyright because is was not derivative enough. 

Are you and other prompters involved and have some level of measurable skill? Sure. That doesn't make you or them an artist. 

In the same way the people who modified the car are not car builders. 

 

I don't have the skill needed to make that prompt. But I'm also not arguing if there's skill involved or not. I'm arguing if the ai art makes the prompter an artist. 

[da na]

6 hours ago, 05032-Mendicant-Bias said:

It takes more steps than you think to generate a decent looking image.

 

In this example I used OpenPose to get the pose from the left image, used ProtogenX34 guided by open pose to generate a 1024x1024 blurry mess with mishaped face and fingers, used GAN4X to upscale with a GAN network, then inpainted and regenerated individual details like hands, face, etc... All while refining positive and negative prompts.

It's not art skills because I can never draw anything even remotely close to the right image, no matter how much time I would choose to learn it. It is certanly creative skills to refine the prompt and get closer to what I had in mind. A D&D high elf mage character.

not to kill the vibe but she has 3 legs, one of which does not connect with her body

[Kinda Bottlenecked]

8 hours ago, 05032-Mendicant-Bias said:

Simple, I like it! The bulk of the creativity was inferred from this database (https://laion.ai/blog/laion-400-open-dataset/)

It's fair value? 0$.
 

I like the AI generating process a lot more than searching for the closest image to what I want on google search. E.g. it was hard getting the crown thingie on her head. I wanted to get a side of her face scarred, but I couldn't do it. The model doesn't know scars. I'll have fun finding ways to do it.

 

I guess this is why the copyright wasn't awarded to any non human artist. Most users of AI generated models rely on picking the perceived "best" image in a seed to arrive at the final generated image. 

Quote

As the Supreme Court has explained, the “author” of a copyrighted work is the one “who has actually formed the picture,” the one who acts as “the inventive or master mind.”

 

As long as this stands, they way that most Art generation models interface with the users do not allow for such composition of the image. From the get go the image you generated uses darker values on the robes of the character which give a nice overall silhouette that separates her from the background. Also aside from the arms that you've position in open pose, the image also has guiding lines that flows to the subject at the center. All these lead to you to the main subject at the center of the frame. 

 

These are what most AI generation models are missing currently. There is no composing of the image, there is nothing that you can decide on. There are parameters you can pick to specify bounds that the model will work in but that is all there is. 

 

[Mark Kaine]

14 hours ago, Fasterthannothing said:

I agree it's dumb and honestly I'm probably gonna be flamed for this but I don't see a big deal with AI art being able to be copywritten. If AI is really generating unique pieces every time then it's going to be unique no matter what.

 

"if this completely not unique piece of work is unique then it should be copyrighted"

 

ie how do you proof its unique... AI cant do anything without outside input,  because its not actually intelligent,  its fake. 

[Arika]

8 hours ago, Mark Kaine said:

"if this completely not unique piece of work is unique then it should be copyrighted"

 

ie how do you proof its unique... AI cant do anything without outside input,  because its not actually intelligent,  its fake. 

What makes you think its not unique? my actual art is very derivative of other works and styles why can i draw something based off something else and copyright it but AI can not?

 

Infact, can i copyright a drawing that has been based entirely on an AI generated picture?

 

[05032-Mendicant-Bias]

10 hours ago, Mel0n. said:

not to kill the vibe but she has 3 legs, one of which does not connect with her body

And this is one of the better generations.

 

The generator has an habit of generating three arms, and make her into a centaur with three to five legs. Sometimes it spawn an extra torso on top of the existing torso. Some images can be nightmare fuel, and /or very funny for how well it blends the extra limbs. Half of the prompt are there to try and remove extra limbs!

The easieset way to erease extra limbs is to inpaint the extra bits, but I couldn't be bothered with it with this image. I already moved on the the next ones to try more things.

11 hours ago, goodtofufriday said:

Inputting commands it not the same no. Using a digital pen, or mouse pointer, acts in the same way a real pen does. Modeling a model in blender is substancially similar to taking real clay and making a model.

That's the point. I cannot use a pen proficiently, but as a programmer I'm built to write commands.

This is a class of tools in the tool box. A different mean to achieve the same end result of arranging pixels the way you want them to.

7 hours ago, Kinda Bottlenecked said:

These are what most AI generation models are missing currently. There is no composing of the image, there is nothing that you can decide on. There are parameters you can pick to specify bounds that the model will work in but that is all there is. 

DallE relies on prompt, more advanced tools like StableDiffusion do allow more control in the way the generation is done, with things like inpainting, outpainting, skeletron, text inversions, styles, etc.... Also chances are that photoshop & co are rushing to add a StableDiffusion/GAN brush in their program. 

10 hours ago, goodtofufriday said:

I don't have the skill needed to make that prompt. But I'm also not arguing if there's skill involved or not. I'm arguing if the ai art makes the prompter an artist. 

I wouldn't mind if there is a special name for prompters that is not artist. Some steps are similar between the two, other steps are unique for artists, and unique for prompters. It's two tasks that are different enough to warrant different names to me. 

I argue that using a AI generator is no different from using a photoshop content aware scissor/brush. The AI generator is just that piece of code taken to the next level.

 

I argue copyright should be evaulated on individual pieces of work, and that if a photograph is worthy of copyright, then using an AI generator as step to get to an output, should yield something that is worthy of copyright as well. The prompter added something. Not brush skills, not any knind of artistic skills maybe, but something creative, in the same way a photographer added something even tough he just pointed and clicked.

[Chiyawa]

Hmm, with AI thing going on, I guess what we need to define is this:

 

AI is a tool, like paint brush, writing equipment, power tools and what not. The copyright still belongs to the creator.

[YellowJersey]

17 hours ago, porina said:

The law is what the law is.

Not necessarily. If that were true, the corpos wouldn't be spending billions every year to lobby politicians to create, change, or scrap laws to their benefit. The law is constantly changing and until there's legislation (preferably with the blessing of the US supreme court) that clearly and unambiguously states whether AI works can be subject to copyright protection there is still a haze of uncertainty surrounding the subject. The current state of the law and the interpretation thereof supports the notion that AI works cannot be subject to copyright protection.

 But the law is what the law is... until it isn't anymore. AI is a disruptive technology and like all disruptive technologies there are going to be those in favour of the status quo and those in favour of expanding the existing law to account for it.