AI art.

Discussion regarding the Outsider webcomic, science, technology and science fiction.

Moderator: Outsider Moderators

User avatar
White
Posts: 73
Joined: Tue Feb 14, 2017 9:26 am

AI art.

Post by White »

Decent AI art is a thing now, in many people's opinions, anyway.

And, though not directly to outsider, I thought that - this being a forum with a major emphasis on art and technology, this might engender some useful discussion.

Mainly, I wanted to listen to your opinions on how this might affect things like art assets for videogames, and perhaps even webcomics, and - if any of you see enough potential in the technology to use it in any capacity for "serious" tasks.

As an example, for those of you that have no idea what I'm talking about. I've posted my attempts to make some Loroi with the technology.

And in case you happen to be reading, Arioch, feel free to add them to the fanart section without credit if you think any of them are relevant enough submissions.
SpoilerShow
Image
SpoilerShow
Image
SpoilerShow
Image
SpoilerShow
Image
SpoilerShow
Image

User avatar
SVlad
Posts: 305
Joined: Thu Aug 27, 2015 5:43 pm
Location: Saint-Petersburg, Russia

Re: AI art.

Post by SVlad »

What was your request strings? And what AI generator did you use?
Outsider in Russian
Image

User avatar
DevilDalek
Posts: 191
Joined: Wed Jul 27, 2011 9:22 pm
Contact:

Re: AI art.

Post by DevilDalek »

Dunno.. my results for
"Epic viking dolphins fighting robot hedgehogs from the moon while eating messy hotdogs"
Was a bit of a mixed bag.

User avatar
White
Posts: 73
Joined: Tue Feb 14, 2017 9:26 am

Re: AI art.

Post by White »

SVlad wrote:
Mon Aug 22, 2022 4:51 pm
What was your request strings? And what AI generator did you use?
The strings were different, but generally followed the format: "Close up of an anime blue-skinned elf girl. Her skin is blue in color. She has dark-blue skin and a short bob of white hair. She stands on the surface of a barren alien planet, the evening stars sparkling in the sky above her. Comic Cover. Trending on Art Station. Extremely detailed. Excellent use of shadows. Painting by. Smooth."

And I was using Dall-E 2.
DevilDalek wrote:
Mon Aug 22, 2022 4:52 pm
Dunno.. my results for
"Epic viking dolphins fighting robot hedgehogs from the moon while eating messy hotdogs"
Was a bit of a mixed bag.
What model are you working with? Try putting fewer elements into a scene. You can make some impressively large scenes if you know how to put different images together.

Bamax
Posts: 1040
Joined: Sat May 22, 2021 11:23 am

Re: AI art.

Post by Bamax »

White wrote:
Mon Aug 22, 2022 4:38 pm
Decent AI art is a thing now, in many people's opinions, anyway.

And, though not directly to outsider, I thought that - this being a forum with a major emphasis on art and technology, this might engender some useful discussion.

Mainly, I wanted to listen to your opinions on how this might affect things like art assets for videogames, and perhaps even webcomics, and - if any of you see enough potential in the technology to use it in any capacity for "serious" tasks.

As an example, for those of you that have no idea what I'm talking about. I've posted my attempts to make some Loroi with the technology.

And in case you happen to be reading, Arioch, feel free to add them to the fanart section without credit if you think any of them are relevant enough submissions.
SpoilerShow
Image
SpoilerShow
Image
SpoilerShow
Image
SpoilerShow
Image
SpoilerShow
Image

Wow... this is nearly what Loroi would look like in real life.

All you need for the finishing touch is to slant their eyes down at a slight angle toward the nose and you would have totally realistic looking Loroi!

Good job! Well done!

Y

User avatar
White
Posts: 73
Joined: Tue Feb 14, 2017 9:26 am

Re: AI art.

Post by White »

Bamax wrote:
Mon Aug 22, 2022 7:31 pm

Wow... this is nearly what Loroi would look like in real life.

All you need for the finishing touch is to slant their eyes down at a slight angle toward the nose and you would have totally realistic looking Loroi!

Good job! Well done!

Y
Eh, the AI did most of the work, haha.

And, you're probably going to have to get someone with an art degree to make those changes.

User avatar
Arioch
Site Admin
Posts: 4486
Joined: Sat Mar 05, 2011 4:19 am
Location: San Jose, CA
Contact:

Re: AI art.

Post by Arioch »

I've seen a few of these kind of posts around recently, and I don't know what specific tools are used to create them (or exactly how they work), but based on what I know from my own software experience as well as using recent "AI" graphics routines in Photoshop, Vizcom, etc., I have the feeling that what they're doing is being oversold. The "AI" routines in Photoshop and Vizcom can spit out some very impressive results, but those results are completely tied to source images that you have to supply. For example, Vizcom is supposed to be a concept design aid that (supposedly) can shade your sketches or generate some new concepts, but it's not "imagining" these things, just applying algorithms on source images. So if you're using some kind of package that takes keywords and spits out images, then I'm pretty sure that there's a step we're not seeing in which the algorithm is either drawing from a library of stock images, or more likely, is doing some kind of web image search and using that as a source. That's probably not a huge concern for an enthusiast, but for a professional content creator, I'd be very concerned about publishing something that may turn out to be a visually obvious rip-off of someone else's work, without even knowing it.

The technology is impressive, but in terms of how useful this sort of thing is to a working artist, I'm not sold. Vizcom can produce a rendered concept of a car if you give it a sketch and some reference photos of cars, but shading a sketch is something that only takes an artist a few minutes, which is generally how long you have to futz with the parameters in Vizcom to get something presentable, and you have to be an artist to create the sketch in the first place. And I had some fun using Photoshop's AI routines to expand the concept paintings of Perrein and Mezan into the larger images that appeared in the comic, but the process required a lot of manual tweaking and overpainting of the output images, and I doubt I saved any time over just doing the whole thing by hand.

User avatar
White
Posts: 73
Joined: Tue Feb 14, 2017 9:26 am

Re: AI art.

Post by White »

Arioch wrote:
Mon Aug 22, 2022 10:11 pm
I've seen a few of these kind of posts around recently, and I don't know what specific tools are used to create them (or exactly how they work), but based on what I know from my own software experience as well as using recent "AI" graphics routines in Photoshop, Vizcom, etc., I have the feeling that what they're doing is being oversold. The "AI" routines in Photoshop and Vizcom can spit out some very impressive results, but those results are completely tied to source images that you have to supply. For example, Vizcom is supposed to be a concept design aid that (supposedly) can shade your sketches or generate some new concepts, but it's not "imagining" these things, just applying algorithms on source images. So if you're using some kind of package that takes keywords and spits out images, then I'm pretty sure that there's a step we're not seeing in which the algorithm is either drawing from a library of stock images, or more likely, is doing some kind of web image search and using that as a source. That's probably not a huge concern for an enthusiast, but for a professional content creator, I'd be very concerned about publishing something that may turn out to be a visually obvious rip-off of someone else's work, without even knowing it.

The technology is impressive, but in terms of how useful this sort of thing is to a working artist, I'm not sold. Vizcom can produce a rendered concept of a car if you give it a sketch and some reference photos of cars, but shading a sketch is something that only takes an artist a few minutes, which is generally how long you have to futz with the parameters in Vizcom to get something presentable, and you have to be an artist to create the sketch in the first place. And I had some fun using Photoshop's AI routines to expand the concept paintings of Perrein and Mezan into the larger images that appeared in the comic, but the process required a lot of manual tweaking and overpainting of the output images, and I doubt I saved any time over just doing the whole thing by hand.
The software I was working with was a new AI model called Dall-E 2. A company called open AI has written a paper about it, but, generally, it's the forerunner in a new field of "diffusion models" that create images from plain text. I won't claim to understand how it works, but it's been described to me as a model that can turn text prompts into vectors, and then turn vectors into images. I think the method is called "Clip guided diffusion" if that means anything to anybody here. I'll leave off speaking any more on the technology, as I'm sure you'd get a lot more information and suffer fewer misunderstanding by looking it up yourselves.

As to whether the software is plagiarizing. In my estimation, I don't think that's the case, considering the outputs I've seen it capable of producing. The company that hosts it seems confident about this fact as well, as they're already handing out monetization rights for any images one creates with it.

You're definitely right about the amount of effort it takes to get something high quality, though. I had to go through dozens of prompts in order to get those outputs, and many of them still have some small issues that could easily be fixed by a human artist. Ironically enough, writing effective prompts for the software is more of an art than a science. Besides which, getting something like a consistent character model or art style has proved... infeasible, to say the least.

Still, I do think it might have some great utility in creating texture assets or one off backgrounds, or in adding extraneous details to this and that and whatnot. I don't know if this would actually be relevant to your workflow since you use 3D models for backgrounds and the like, but I do think it's utility is on the order of something that can make some quality of life improvements for even professional artists.

User avatar
Arioch
Site Admin
Posts: 4486
Joined: Sat Mar 05, 2011 4:19 am
Location: San Jose, CA
Contact:

Re: AI art.

Post by Arioch »

In order to have a true AI that could create an original illustration of a human subject (that is, not one based on existing images) from keywords, that AI would have to have an adult human's understanding of natural language, human anatomy, fashion, culture, not to mention the physics of lighting (which is actually the easy part). There's literally no way that's what's happening. Current AI is impressive, but not that impressive.

Machine learning techniques produce effective output by ingesting large amounts of data and looking for patterns. If you ask for a picture of an elf, the only way it can supply this is by ingesting a huge number of pictures identified as elves, identifying patterns and then synthesizing a simulacrum.

Taking a quick glance at the DALL-E 2 website, that's exactly what it's doing: it's searching for source images based on the keywords you specify, analyzing them for patterns, and then using those patterns as a template to create a new image. The results are surprising and impressive, but it's not creating anything original; it can't create something for which references do not exist. And despite what the description misleadingly says, it doesn't "learn" what an elf is; its output is only as good as the keywords and source images are. The more source images, the less the output will be like any individual image... but since that stage is being hidden from us, we don't know how many images were used as source for a particular render, or which ones.

The fact that you used "Artstation" as a keyword is curious and perhaps telling.

I did a few Google Image searches on a few of your renders (which is more or less the same AI process in reverse), and it didn't find any exact matches... but there are very specific details like the costumes which had to have been pulled from somewhere.

User avatar
White
Posts: 73
Joined: Tue Feb 14, 2017 9:26 am

Re: AI art.

Post by White »

I have a feeling we actually agree on the capabilities of the AI, but perhaps have assigned different words to the same thing.

For instance, it is true that Dall-e is trained on a variety of source images. However, I don't think it searches for those images when you use it. Theoretically, you could download the model onto your own hard drive, without connection to a database of images, and you could use it perfectly well assuming the weights on the neural net are set

So, you're definitely correct when you say that it's analyzing source images, however, it's not doing that in real time. It's already "digested" the hundreds of millions of images from it's training data, and it - along with other diffusion models - seem once trained able to operate without a connection to their training dataset.

I don't know if this makes a difference to your argument, as, at the end of the day, it is still analyzing patterns from word-image pairs, but I do think that should give it a slightly higher level precedence when compared to a photobasher.

And, again, you're absolutely right when you say Dall-E doesn't demonstrate understanding the same way a human does. It's said that Dall-E returns the most "probable" result for a set of words, not the "best". It does essentially function like a search engine, hence why terms like "Trending on Art Station" or "Very beautiful" can end up affecting the output so much.

However, I do still believe that the program doesn't plagiarize any of it's results, and thus is viable for commercial use.

And trust, as someone who uses it, I'm more aware than most of how very, very limited it is when compared to a human artist and the things they're called upon to do as a part of their daily jobs. Hence why I was only suggesting things like backgrounds and texture assets as primary use cases.

TLDR: I don't claim the AI "understands" the prompt. However, I don't think understanding is required in order to get original content. Hence why I'm sure that the program isn't plagiarizing.

User avatar
Arioch
Site Admin
Posts: 4486
Joined: Sat Mar 05, 2011 4:19 am
Location: San Jose, CA
Contact:

Re: AI art.

Post by Arioch »

White wrote:
Mon Aug 22, 2022 11:50 pm
So, you're definitely correct when you say that it's analyzing source images, however, it's not doing that in real time. It's already "digested" the hundreds of millions of images from it's training data, and it - along with other diffusion models - seem once trained able to operate without a connection to their training dataset.

I don't know if this makes a difference to your argument, as, at the end of the day, it is still analyzing patterns from word-image pairs, but I do think that should give it a slightly higher level precedence when compared to a photobasher.
I very much doubt that they're doing image searches in real time; the site mentions that there are "automatic and manual" controls in place to try to prevent pornographic or graphically violent images from being used as source, so it's more likely that they are collected in batches and preprocessed. But the collection method isn't really relevant to the argument; the only place they could be sourcing images from is the Web.
White wrote:
Mon Aug 22, 2022 11:50 pm
It does essentially function like a search engine, hence why terms like "Trending on Art Station" or "Very beautiful" can end up affecting the output so much.
That should be setting alarm bells off in your head.
White wrote:
Mon Aug 22, 2022 11:50 pm
However, I do still believe that the program doesn't plagiarize any of it's results, and thus is viable for commercial use.
One of the examples shown on their website is of a famous Vermeer painting that DALL-E has made alternate versions of. None of them are exact copies of the original, but they are clearly done in Vermeer's style and they're obviously based on that particular painting. Now, if Johannes Vermeer was still alive, and someone was trying to commercially sell these DALL-E renders... do you think he'd be okay with that?

User avatar
White
Posts: 73
Joined: Tue Feb 14, 2017 9:26 am

Re: AI art.

Post by White »

Arioch wrote:
Tue Aug 23, 2022 12:04 am
I very much doubt that they're doing image searches in real time; the site mentions that there are "automatic and manual" controls in place to try to prevent pornographic or graphically violent images from being used as source, so it's more likely that they are collected in batches and preprocessed. But the collection method isn't really relevant to the argument; the only place they could be sourcing images from is the Web.
Again, we may be talking past one another on this. But I have no disagreements with the fact that the training data comes from the web.

However, I would disagree with the idea that a "search" makes up any part of creating an output. I think once the model is trained, it just works with whatever correlations it managed to learn by studying the data.

I'd consider it more akin to the AI taking inspiration from a vast amount of training data rather than cordoning off a specific section and plagiarizing.

So, when one asks for a chair made out of broccoli, I don't think the AI is searching a database for broccoli and again for chair and then trying to make an image out of it's source material. I think the resulting image will be the result of whatever effect the entirety of the training data had on the neural net weights. This makes it incredibly unlikely that it will plagiarize, I would think.
That should be setting alarm bells off in your head.
Again, I don't agree with the assertion that originality requires a human level understanding of motives. That old google-art bot that just made fractal dogs probably had less "understanding" than Dall-E. Yet I don't think there's a single human on earth that would be ready to start a copyright case over it's work.
One of the examples shown on their website is of a famous Vermeer painting that DALL-E has made alternate versions of. None of them are exact copies of the original, but they are clearly done in Vermeer's style and they're obviously based on that particular painting. Now, if Johannes Vermeer was still alive, and someone was trying to commercially sell one of these DALL-E renders... do you think he'd be okay with that?
If you asked a human for particular Vermeer painting in the style Vermeer, they'd probably also be plagiarizing to get you wanted.

I'm not claiming that the AI is incapable of plagiarizing. If you ask it to plagiarize someone explicitly, it probably will. However, I just doubt that any innocuous request, like, for example, a texture pack, is likely to plagiarize anything. In fact, the fact that none of the Vermeer alternates were exact copies of the original probably hints at the fact that AI isn't doing anything shady when creating it's outputs, like ripping off lesser known artists.

Of course, if people decide to use Dall-E to plagirize Vermeer or anyone else's style. That would be plagirism plainly. But, that already happens with human artists, because humans plagirise, and some might use an AI to do it, but I don't it's fair to use this as proof that the AI is incapable of creating non-plagirized outputs.

And, I'll admit right now that there is a lot of contextual stickyness when it comes to the AI's output. I've got some stuff written up on this if you want to see it, but it will generally import a lot of implicit values along with whatever you explicitly ask for.

For instance, if you ask for a painting by someone from the middle ages, it's difficult to get figures that are wearing non-medieval clothing.

Again, this is a part of the charm, and what allows it to create such detailed outputs with short prompts. However, this shouldn't cause any more accidental plagiarism than humans alone are guilty of, as a cascade of related matters ends up making two works look more than similar than pure probability would dictate.

Again, none of this is to say that diffusion models are on par with, or will be on par with human artists. Understanding is a key point in creating consistent charachters that are depicted consistently from multiple angels and in relation to various objects. (Again, I have a lot of detailed examples where the AI does strange things with object relation, if you want some stories."

For example, I remember there was an AI that managed to "generally" become superhuman at every Atari game they threw at it except the temple of montezuma. This being because games like starfighter were basically reaction time based, and Temple of montezuma required a player to navigate a temple and collect keys and solve puzzles and head towards various goals and doorways that needed to be completed in a certain order... you get the point.

I still stick to my point that a human level of understanding isn't required for the creation of a non-plagirising AI, and this would especially be the case when it comes to things like backgrounds and texture assets and the like.

User avatar
Arioch
Site Admin
Posts: 4486
Joined: Sat Mar 05, 2011 4:19 am
Location: San Jose, CA
Contact:

Re: AI art.

Post by Arioch »

If you ask for a Vermeer, you know who you're ripping off. If you don't specify an artist, you're still ripping off someone; you just don't know who. And that's one of the big problems I have with a system like this: it's a black box. We can't see the reference images used, and so we don't know to what extent the final product is similar to any or all of them. I don't have the time or inclination to comb Artstation to find the pieces of artwork that those costume details in your first four images were stolen from, but make no mistake: they were stolen from someone. Those costume elements were not designed by DALL-E; it doesn't have that capability. It can only reproduce something it has seen before. The wording they use in regard to upload policy leaves me with the impression that source images might be manually uploaded by users. But this cannot guarantee that those uploading them have the rights to grant consent for use; there's no way this could be manually checked for the volume of images that they must be using. Maybe it's the case that all of these scifi costume elements were culled from illustrations that were uploaded into the system by the artists that created them, but that seems extremely unlikely to me.

This is probably going to be an issue where new law will be required, as I expect the current laws didn't imagine a system like this, but images culled from the internet are not public domain; they are still owned by someone. OpenAI has not secured the rights of all of the images on the internet to feed to its algorithm, because that would be impossible. Looking at their content policy, the only use guidelines I can find are mostly concerned with the deceptive use of altered photos; there's nothing there that addresses the copyright of design elements. All that it says is that they allow you to post DALL-E images on social media, with the caveat "You must clearly indicate that images are AI-generated - or which portions of them are - by attributing to OpenAI when sharing, whether in public or private." Nowhere (that I can find) does it say that these images are legal to use in commercial projects. Since you're using the software, maybe you know more about their content use guidelines.

User avatar
White
Posts: 73
Joined: Tue Feb 14, 2017 9:26 am

Re: AI art.

Post by White »

Again, it's outputs to every random idea has convinced me that it's capable of stringing enough unique ideas, to make something original enough not to get sued, and that it's definitely "painting" things from scratch rather than modifying existing images

I'll grant of course that this is mainly a hunch based on personal experience, and I'll admit that hunch mainly applies to the subjects of the paintings and not the styles, which are harder to test for, but... considering I had no intention of using it for commercial use anyway, It's not a question that keeps me up at night.

As to the similarity of the outfits in the first four images. I should note tha there is something in the AI that makes it give similar results to the same prompt. As I mentioned with the contextual stickiness earlier, it has some boundary of relating similarly related objects together like clothing and time period. And while it may have certainly taken inspiration from the amalgamation of posters in it's data set, I don't think it's plagiarized a single one.

Of course, as you said, It wouldn't be fair to ask you to comb through the internet to prove your point otherwise, so I suppose we ought to agree to stop this particular argument here. If the AI really is plagiarizing... well it shouldn't be able to hide that fact for long considering they've just released it. Of course, if you had any test prompts that might give us hints as to it's workings, I'd be happy to run them.

User avatar
spacewhale
Posts: 123
Joined: Fri Jul 24, 2020 7:08 am

Re: AI art.

Post by spacewhale »

It'd be cool if they'd just link all the assets they've trained the AI off of.

User avatar
Werra
Posts: 840
Joined: Wed Jun 06, 2018 8:27 pm

Re: AI art.

Post by Werra »

I for one welcome a program that puts the art in artificial intelligence. That these AIs brute force creativity doesn't bother me either. In a way, human brains do the same. If you draw a space elf OC you are not plagiarizing a specific image. Sure. But we still have a framework of what a 'space uniform' looks like which is formed by other images. Given enough example pictures to draw from a decent AI should be able to produce an original image. Whether it counts as art or not depends on the product.

While an artist might disagree, Dall-E's output seems original to me. Because we can use Dall-E to generate images of real objects, like pizza. Did Dall-E plagiarize olives on pizza or did it merely correctly identify trends in the data? Somewhat uninspired maybe, but Dall-E strikes me as a graphic design workhorse. It won't replace the passionate artist who pours his mania and cheap wine into his works. But it or follow-up products will be serious competition to people putting out standard, safe work. The poor shmucks who make the box art for Microsoft Office.

Or I am totally wrong and the inspired weirdos will be made obsolete and the office drones will find new purpose touching up AI generated images. Dall-E has put out genuine artistic works already. If you can find it check out '9/11 gender reveal', the perfect synthesis of a celebration and human tragedy.

Bamax
Posts: 1040
Joined: Sat May 22, 2021 11:23 am

Re: AI art.

Post by Bamax »

Werra wrote:
Tue Aug 23, 2022 9:59 pm
I for one welcome a program that puts the art in artificial intelligence. That these AIs brute force creativity doesn't bother me either. In a way, human brains do the same. If you draw a space elf OC you are not plagiarizing a specific image. Sure. But we still have a framework of what a 'space uniform' looks like which is formed by other images. Given enough example pictures to draw from a decent AI should be able to produce an original image. Whether it counts as art or not depends on the product.

While an artist might disagree, Dall-E's output seems original to me. Because we can use Dall-E to generate images of real objects, like pizza. Did Dall-E plagiarize olives on pizza or did it merely correctly identify trends in the data? Somewhat uninspired maybe, but Dall-E strikes me as a graphic design workhorse. It won't replace the passionate artist who pours his mania and cheap wine into his works. But it or follow-up products will be serious competition to people putting out standard, safe work. The poor shmucks who make the box art for Microsoft Office.

Or I am totally wrong and the inspired weirdos will be made obsolete and the office drones will find new purpose touching up AI generated images. Dall-E has put out genuine artistic works already. If you can find it check out '9/11 gender reveal', the perfect synthesis of a celebration and human tragedy.

I think any artist of skill won't like the idea of equally skilled AI artists. Not without some major changes to law to prevent loss of income.


Imagine if you will an AI that could easily reproduce all the art in Outsider. A rich tech company then hires the most talented writers they can find and then make a scifi comic that uses many of the same concepts as Outsider... verging on plagerism (Babylon 5/DS9 debacle).

The new comic outshines Outsider simply because they have more money to throw at advertising.


Would Arioch like this? I think not so much.

User avatar
elorran
Posts: 69
Joined: Thu Mar 10, 2011 9:31 pm

Re: AI art.

Post by elorran »

AI learning and application isn't going away, and its only going to get more efficient and wide spread with time. I personally like that AI art is becoming a thing, but don't like that its going it impact the traditional artists themselves who make a living off commission work.
AI won't replace a tradition artist any time soon, but its already an open pandora's box for misrepresentation and scams. Hopefully (but doubtfully) this won't lead to a tread of abuse of traditional artists or stealing their work (NFT thieves are already doing this en-masse sadly).

User avatar
Werra
Posts: 840
Joined: Wed Jun 06, 2018 8:27 pm

Re: AI art.

Post by Werra »

Technology has revolutionized lots of fields already. We've been losing jobs to industrialisation for centuries. Why would art be safe?

User avatar
White
Posts: 73
Joined: Tue Feb 14, 2017 9:26 am

Re: AI art.

Post by White »

Bamax wrote:
Wed Aug 24, 2022 12:41 am
Werra wrote:
Tue Aug 23, 2022 9:59 pm
I for one welcome a program that puts the art in artificial intelligence. That these AIs brute force creativity doesn't bother me either. In a way, human brains do the same. If you draw a space elf OC you are not plagiarizing a specific image. Sure. But we still have a framework of what a 'space uniform' looks like which is formed by other images. Given enough example pictures to draw from a decent AI should be able to produce an original image. Whether it counts as art or not depends on the product.

While an artist might disagree, Dall-E's output seems original to me. Because we can use Dall-E to generate images of real objects, like pizza. Did Dall-E plagiarize olives on pizza or did it merely correctly identify trends in the data? Somewhat uninspired maybe, but Dall-E strikes me as a graphic design workhorse. It won't replace the passionate artist who pours his mania and cheap wine into his works. But it or follow-up products will be serious competition to people putting out standard, safe work. The poor shmucks who make the box art for Microsoft Office.

Or I am totally wrong and the inspired weirdos will be made obsolete and the office drones will find new purpose touching up AI generated images. Dall-E has put out genuine artistic works already. If you can find it check out '9/11 gender reveal', the perfect synthesis of a celebration and human tragedy.

I think any artist of skill won't like the idea of equally skilled AI artists. Not without some major changes to law to prevent loss of income.


Imagine if you will an AI that could easily reproduce all the art in Outsider. A rich tech company then hires the most talented writers they can find and then make a scifi comic that uses many of the same concepts as Outsider... verging on plagerism (Babylon 5/DS9 debacle).

The new comic outshines Outsider simply because they have more money to throw at advertising.


Would Arioch like this? I think not so much.

Not sure it's that simple.

Corporations can already hire talented writers. And, as far as the art work goes, they have already proven to be able to hire legions of artists that can push out a consistently good products. See: Marvel comics.

Yet, I'm sure you'll agree that the average marvel comic, for a combination of marketing and institutional reasons, isn't really that much better than the average webcomic. And, frankly, the really great marvel comics generally are hindered from being too creative by the aforementioned infrastructure inherent to the comics industry.

See how Japan's model of buying original works made by individual artists is financially and (perhaps artistically) better than US comics by a wide margin.

The point to make is: I doubt a lack of resources to buy art with has been what's been stopping companies from taking over this space. If anything, /ai art would be decreasing the competitive advantage that companies used to have (a monopoly on expensive art).

And this is assuming that we have an AI that can easily reproduce ART that's good enough to go in a comics panel, or that diffusion modes are equipped to do that. I'm not saying that it's impossible, but I don't see that possibility has been adequately demonstrated by the models on offer.

Post Reply