Background Pony #0306
FYI, this is being linked to in an AI thread on /mlp/
About AI art theft...
Background Pony #A2CE
Background Pony #743C
The more I go through those comments coming from “artists”, the more I’m convinced the AI generated art is a great opportunity to get something beautiful from something that doesn’t seethe or rage in public places - just like some people on this forum. I’d even prefer to pay a fee to an AI art generating site to get a commission done, than paying such and indivual who post some angry nonsense on the Internet.
Another reason to support AI generators is: most of them won’t complain about the theme of a commission, just like some artists do. “Oh, I don’t draw X”, “It’s not my fetish”, “commissions closed, I’ll open eventually in 3 years if you praise me and give attention for long enough~”. AI generators won’t do it and it’s a huge advantage compared to whining “artists”.
Those people who say “let everyone know that WE DO NOT WELCOME AI GENERATED STUFF HERE”, just a small advice - suck it up, nobody cares, the majority of people is not a part of your “we”.
Another reason to support AI generators is: most of them won’t complain about the theme of a commission, just like some artists do. “Oh, I don’t draw X”, “It’s not my fetish”, “commissions closed, I’ll open eventually in 3 years if you praise me and give attention for long enough~”. AI generators won’t do it and it’s a huge advantage compared to whining “artists”.
Those people who say “let everyone know that WE DO NOT WELCOME AI GENERATED STUFF HERE”, just a small advice - suck it up, nobody cares, the majority of people is not a part of your “we”.
Background Pony #A2CE
@Background Pony #743C
And another obvious 4chan “visitor”.
And another obvious 4chan “visitor”.
May as well just lock the thread now, there’s going to be a number of people coming in that hate this site, hate artists, or both, and will say whatever they think will make either the angriest (even if that is just them being themselves) and make whatever discussion may have been possible before impossible.
Cocaine 
Princess Coke
So many callous and hateful words towards people who worked for the fandom the most. I think my point is proven pretty well. AI crap already emboldened hateful people to directly attack real artists just because they feel entitled to free labor from them. Artists not ur slaves guys. They deserve every penny for their hard work.
Ciaran 
Senior Moderator
友情は魔法だ
@Background Pony #A2CE
I have to admit, there is some merit to that suggestion.
I have to admit, there is some merit to that suggestion.
Despite the title of the thread, no one (posting in this thread) has provided or reported a single example of actual art theft.
I’ve described what we would do if someone did report actual art theft, but some people think that’s gaslighting. And so far me asking what it is that people think should be changed in our existing policies has only resulted in personal attacks and demands that we “ban everything” AI related.
Which, as most people know, is a lot of stuff. The MLP show itself is heavily dependent on what is ostensibly AI for tweens, filters, textures, etc.
So, I’m going to go watch ThisPonyDoesNotExist for awhile and contemplate locking this.
Friendly reminder to artists in this thread - with actual art on this site - that they should report art theft for Rule #1. That’s not allowed here, regardless of who or what is violating your copyright.
@Cocaine
Report the art theft you see happening. Please do it now.
Report the art theft you see happening. Please do it now.
hateful people to directly attack real artists
If that is happening on this site, report it. Stop complaining that other people should do something about it or that banning all AI generated content would somehow magically prevent it - if you see it happening, report it.
d1ckbitch 
Doughnuts Worshipper
Before the thread is locked, here’s my suggestion : why not actually assessing the technology itself, and see what it is capable of, and how cherrypicked the few daily results posted on DB are?
AI seems really good at lighting, but obtaining an actually half-good looking picture is time-consuming and probably would ask some manual fixing, aka being profilient with art tools…
Did it myself, and seeing the huge amount of horrible results I’ve produced yet was kind of comforting ( ͡° ͜ʖ ͡°)
Cocaine
Princess Coke
@d1ckbitch
It have nothing to do with what I said. I assume ai pics will be indistingvasheble from real artist artwork at some point. The ai isn’t the problem. Use of ai and what culture it cutivates is the problem.
It have nothing to do with what I said. I assume ai pics will be indistingvasheble from real artist artwork at some point. The ai isn’t the problem. Use of ai and what culture it cutivates is the problem.
@Ciaran
I just pointed out very explicit result of what ai generators cultivate. Hostility is very real and right in front of our eyes.
I also had an impression that site staff could do something for artists and somehow tweak the rules to accomodate to the changing situation. Current ones isn’t really suitable for the present situation. Just as u said no one was able to use rules to report art theft commited by the bot users.
Reporting rude people doesn’t really do much, while banning ai generated pics would ostracized people like that and help the artists.
Anyway I’m not gonna continue, this thread was quite a waste of time.
I just pointed out very explicit result of what ai generators cultivate. Hostility is very real and right in front of our eyes.
I also had an impression that site staff could do something for artists and somehow tweak the rules to accomodate to the changing situation. Current ones isn’t really suitable for the present situation. Just as u said no one was able to use rules to report art theft commited by the bot users.
Reporting rude people doesn’t really do much, while banning ai generated pics would ostracized people like that and help the artists.
Anyway I’m not gonna continue, this thread was quite a waste of time.
Background Pony #AB6B
@Background Pony #743C
AI art will never be the same as hand drawn. Also why do you care if some artists won’t draw what you want? Everyone has different tastes. I also love how you generalize every artist into one “whiny” complaint.
AI art will never be the same as hand drawn. Also why do you care if some artists won’t draw what you want? Everyone has different tastes. I also love how you generalize every artist into one “whiny” complaint.
Ciaran
Senior Moderator
友情は魔法だ
Hostility is very real and right in front of our eyes.
Believe me, we’ve been dealing with hostility around machine generated images since people were using Flash to make OCs.
I also had an impression that site staff could do something for artists and somehow tweak the rules to accomodate to the changing situation
As announced by one of our admins, we have added machine learning generated to the filter lists of all of site’s default filters. And, in general, such images aren’t permitted in most of our community events.
In practice, where an AI generated work has been shown to be a copyright violation (or theft) - whether that is a recreation of paywalled content, or effectively a “trace” of an original work, we have treated it as a Rule #1 violation of the original artists work.
Also, floods of such images, such as from sources like ThereIsNoSuchPony, have been treated as spam - unlike art made by a Human artist, we don’t generally accept “alternate versions” of machine generated images. Additionally, and also unlike Human made art, machine made images have had to have at least a minimum level of subjective quality.
And, as stated previously, we do not extend copyright protection to machine made images.
They always have been, and continue to be, a separate class of images. And in several ways the people who make the machine made images are not given the same protections or rights as human artists.
To quote an admin of this site;
We feel like artwork made by humans should be the forefront of the site.
Anyway … you continue:
Current ones isn’t really suitable for the present situation
You have not made a case for this. I’ve shown multiple examples of how existing or recently changed rules are suitable. But you have not shown how they are not or provided any examples of them failing.
Just as u said no one was able to use rules to report art theft commited by the bot users.
I said you have not, nor has anyone in this thread who has claimed “theft” is happening.
Others have reported machine-generated derivative images which have been taken down for Rule #1 because they violated an artist’s copyright. We’ve been dealing with those kinds of problems for years, and in every case I’ve been involved with the human artist won.
You keep claiming theft, but you have not shown any.
Our rules aren’t failing. You are failing to use them - all you have to do is report any instances of the theft you claim is happening.
If you are aware of actual theft - report it.
Reporting rude people doesn’t really do much, while banning ai generated pics would ostracized people like that and help the artists.
People using machine learning have actually been some of the more polite people in this discussion, so your argument is not very persuasive.
And you have not made a case that banning AI generated images would help artists.
AI generated images can be beautiful, and inspirational, and can help human artists expand their works.
AI generated art is showing us palates that humans would not normally consider. Layers of details that no human would normally incorporate. And cubist monstrosities that boarder on the horrific yet also show us things about the subjects - ponies - that are both the stuff of nightmares and yet also dreams.
AI generated art can be amazing.
Personally, I think that some people making AI generated art are creating unique and distinctive works that should be copyrightable.
But it’s like Warhol’s painting of Prince.
Who actually owns it? Goldfield whose image inspired Warhol? Warhol who painted it? Or Prince who originally created … Prince?
Right now we are just in yet another controversy about Machine Generated Art.
It’s not the first, it won’t be the last.
But right now, I do not see how banning images because of how they were made, or ostracizing the people who make it, will help anyone.
Background Pony #743C
Personally as an artist who created a lot of pony pics past the last 10 years I will be more than happy to find AI generated art based on my work. The more high quality, cute ponies we get, the better!
If someone feels like AI is stealing their audience/commissioners/money, then oh well - it’s not the first time in the history of mankind when one profession is taken over by the process of automation and machines. Of course it doesn’t mean artists will go extinct, but I believe commissions of digital art for money may die out at some point. Like I said before - I’d prefer to pay for an extremely quickly generated art to get a few iterations of what I’d like to get. No more waiting for months for a slot, no more dealing with personal problems of artists, no more feeling guilty for asking to tweak this or that, or change everything entirely, no more paying tons of money for a single picture.
If someone feels like AI is stealing their audience/commissioners/money, then oh well - it’s not the first time in the history of mankind when one profession is taken over by the process of automation and machines. Of course it doesn’t mean artists will go extinct, but I believe commissions of digital art for money may die out at some point. Like I said before - I’d prefer to pay for an extremely quickly generated art to get a few iterations of what I’d like to get. No more waiting for months for a slot, no more dealing with personal problems of artists, no more feeling guilty for asking to tweak this or that, or change everything entirely, no more paying tons of money for a single picture.
@Cocaine
So far you are quite rude and aggressive, I did nothing to you to make you so angry and desperate. If me (or other people) having a different opinion than you makes you so angry, then maybe you should take a break, grab some coffee, go for a walk and re-consider your anger. In this world having different opinions is a part of every day life and you won’t change it by accusing people of hate, bullying or whatever you want to call it to lock this thread. I don’t have the same opinion as you, yet I don’t demand to lock this thread or to ban anyone.
So far you are quite rude and aggressive, I did nothing to you to make you so angry and desperate. If me (or other people) having a different opinion than you makes you so angry, then maybe you should take a break, grab some coffee, go for a walk and re-consider your anger. In this world having different opinions is a part of every day life and you won’t change it by accusing people of hate, bullying or whatever you want to call it to lock this thread. I don’t have the same opinion as you, yet I don’t demand to lock this thread or to ban anyone.
Background Pony #65C1
I’m hiding as a background pony because of fear of being attacked, but since this is an important topic that can determine the future of the site, I feel like have to add my two cents and state that not every artist is opposed to AI generated art. Most artists I’ve asked aren’t actually opposed to the idea, given what’s currently in place on the site.
The biggest concern was actually that they believe that AI generated art should be hidden off sites that were designed for human-made art, but like I said, that sort of system is already in place.
None of them had concerns of “art theft”, and I too have no issues when it comes to someone using my artwork in a dataset to train an AI.
Parallel Black 
@Background Pony #743C
As far as I can tell Cocaine only made one genuine insult throughout this entire thread, not counting referring to AI art as “AI crap”, but that’s not exactly a big deal. It isn’t worth people piling onto them about. Besides, I’m sure you can tolerate a little vitriol considering its their entire profession that’s potentially under threat from this new thing. The onus is on the AI side of the argument to make the case for why it should be tolerated, not on the real artists, who have filled this site for over a decade now, to make space for AI stuff without complaint.
As far as I can tell Cocaine only made one genuine insult throughout this entire thread, not counting referring to AI art as “AI crap”, but that’s not exactly a big deal. It isn’t worth people piling onto them about. Besides, I’m sure you can tolerate a little vitriol considering its their entire profession that’s potentially under threat from this new thing. The onus is on the AI side of the argument to make the case for why it should be tolerated, not on the real artists, who have filled this site for over a decade now, to make space for AI stuff without complaint.
Background Pony #1F39
I’d just like to point out that discussions like these happened over a century ago when cameras started being a thing. A lot of people truly believed it was the end of all art, that there was no skill or artistry involved in taking pictures, that the blackboxes should be destroyed and those taking or exhibiting photographs should be shunned etc.And not only did people continue to create art, but the number of people doing so exploded. And ever since tablets were invented, along with digital art editing programs, it’s exploded even further. Along with the agreement that photography is art in itself, just a different type of art than using a pencil or paintbrush.The invention of cars also didn’t lead the the extinction of horses or horse riding; it made travel more accessible to everyone, and elevated horse riding from something common to something of high value. The invention of cinema didn’t run theatres out of business, it made entertainment more accessible, created new industries, and elevated the quality of theatre even further.And as for coding being art, outside of what the code is going to be used for, it definitely can be. Sure, for most programmers it’s just a mechanical task, like a McDonald’s cook preparing a burger; but a chef at a 3-star Michelin restaurant is definitely creating art when they prepare a meal, right? Both are effectively doing the same thing in the end.
Background Pony #65C1
@Parallel Black
You need not be concerned about AI replacing anyone; there will always be imperfections in the AI; people are merely fine-tuning and presenting the best of the best, the majority of which were manually tweaked in software afterward. Yes, with a little patience and hundreds of generations, you can get lucky and have something with acceptable anatomy that just barely misses the uncanny mark, but anything more than a “portrait” image of a pony face will look wonky with horrible anatomy.
You need not be concerned about AI replacing anyone; there will always be imperfections in the AI; people are merely fine-tuning and presenting the best of the best, the majority of which were manually tweaked in software afterward. Yes, with a little patience and hundreds of generations, you can get lucky and have something with acceptable anatomy that just barely misses the uncanny mark, but anything more than a “portrait” image of a pony face will look wonky with horrible anatomy.
“Pony on rollerblades”, “pony running through the fields”, “pony lying on bed”, all of these the AI is incapable of. Anything slightly complex and everything goes haywire. This was generated with the same AI as the one above.
Exedrus 
The MLP show itself is heavily dependent on what is ostensibly AI for tweens, filters, textures, etc.
I know you weren’t the one who added this point to the thread. But I do have to push back against it.
Classifying basic computer graphics algorithms like tweens or simple image filters in the same league as machine learning is not common in computer graphics. Though many of the algorithms that existed in ToonBoom/Flash/Photoshop/etc before the current AI-boom do have minor similarities to cutting-edge AI (like Stable Diffusion), the two are generally very different. The key difference is that modern AI is deeply data-driven. At their core, the AI algorithms are very sophisticated forms of pattern recognition. They’re like fitting a line to a cluster of points except the points are images and the line is a really complicated shape with 1000+ dimensions it can span across. This means that modern machine learning/AI require large amounts of data to train off of. Without it, the algorithm is worthless; it can’t build a “line”/“shape”/model to describe the patterns. The same is not true of much simpler computer graphics techniques like tweening, blurring, applying layer “modes” like multiply/screen/dodge/burn, and many of the filters used in image/animation authoring software. Those use the same basic math, but generally are pretty simple in what they do. Which means they didn’t need to train on images to function. I don’t remember seeing anything in MLP that would’ve required AI-caliber algorithms; it all looked like pretty standard 2D rendering for the day.
I don’t know that that will affect any of the debate in this thread, but I don’t want people getting the wrong idea of how widespread AI is.
LemonDrop 
C++ Crazed
@Exedrus
That is true but one could also argue the humans who authored those algorithms did the job machine learning is doing now. Humans with a task of creating some sort of heuristic algorithm that best fits real world cases will often look at data and try to devise something that best matches what sort of output they want. So really their brain is doing the analysis and validation of how the output looks with incremental changes to it to improve the performance just like a neural network’s training process would do (or rather a genetic algorithm is probably a more apt comparison there).
That is true but one could also argue the humans who authored those algorithms did the job machine learning is doing now. Humans with a task of creating some sort of heuristic algorithm that best fits real world cases will often look at data and try to devise something that best matches what sort of output they want. So really their brain is doing the analysis and validation of how the output looks with incremental changes to it to improve the performance just like a neural network’s training process would do (or rather a genetic algorithm is probably a more apt comparison there).
Often times even in the older ways of developing algorithms people would use more automated data-driven approaches which are quite similar to what modern machine learning does. Some examples include regressions (for fitting functions to data), bayesian inference (for like say e-mail spam prevention) and other more fancy techniques. Machine learning really is just the logical next step of this data driven programming approach, and that to begin with was just done to eliminate the guesstimation involved in human-created algorithms. ML in general is also not “new”, it has just evolved to this point from simple perceptrons many decades ago which pretty much functioned like linear regressions to the complex deep convolutional networks we have today (as well as hardware advancements making networks of these sizes possible to evaluate and start producing impressive results from).
Of course this mostly applies when working with fuzzy data or things with no simple solution, obviously more precise things often have more discrete solutions that can be derived with math and implemented exactly in which case ML is not needed to begin with (since an exact solution can be found), but yeah I think you get what I’m saying.
Ciaran
Senior Moderator
友情は魔法だ
Good points, but we have had people also complaining about a wide variety of machine-generated artwork, including everything from thisponydoesnotexist to kisekae.
Kind of in the same way we used to get complaints about vector artists “not really being artists”. Or people complaining that art from video games was “not really art”. O people complaining that an animation not 100% made of key frames is … you guessed it; “not really art”.
Seems like we keep passing through phases of what the new “not really art” is.
And we do seem to be getting a new “this should be banned and anyone posting it should be ostracized” more often, year by year.
I mean, this year art has been completely ruined for all time forever for everyone by “thisponydoesnotexist”, “calarts”, and now the new stable diffusion. And the year isn’t over.
Anyway, I’ve been trying to find some examples of AI images that were taken down for violating an artists copyright, but it’s tricky - if it was deleted was because it violated a fan artists copyright, I need that artists permission to even use it as an example.
The best example I can think of is one where someone used AI to make a nude version of an image to recreate a Patreon Only image from the publicly released image.
As others have described, even though the person who did it used AI to create the Patreon Exclusive … “DeepFake” let’s call it, for lack of a better term … they still had to do a ton of work.
Like this image >>2959654 (NSFW) they had to first basically create a collage of the image they wanted generated, as described here (also NSFW). And then work the results and resubmit it over and over and over until it worked.
At that point, they almost could have done it themselves in PhotoShop.
So, for that example at least, their commitment to doing it with AI was more of a fetish, or a desire to use the tool because they trusted it more than their own art skills.
It’s like if I help with an image for the collab, I use Photoshop even if Illustrator is the right tool, just because that’s what I’m most used to.
My sense right now is that AI, and especially some of the newest implementations of it, will be kind of like that in a few years; just an artistic choice.
Exedrus
The underlying idea of a lot of machine learning (perceptrons) was definitely imagined a long time ago. Building a human-like brain in a computer is inherently fascinating. And I’d bet statisticians have always been hanging around the CS space testing probabilistic algorithms. But it wasn’t until the internet gave computer scientists giant mountains of raw data that they started to be able to actually train/test these large networks. Even now, CNNs exist as a shortcut to avoid larger nets that would require more time/data to train up. Having lots of data is the key to “modern” AI that everyone is thinking about.
While older graphics work definitely involved people messing with image data to get a desired effect, it wasn’t on the same scale. They’d use a hand full of images to show the effects, not thousands like what modern ML tries to learn from. From a philosophical angle it’s got some interesting similarities, but practically it’s pretty easy to tell the two apart. A modern ML implementation easily has thousands of lines of code and a brick of model data. A simple image authoring tool like Gaussian blur is like 50 lines of code with no model data.
There are probably common roots between CG algorithms and computer vision. Like experimenting with convolution/kernels to do edge detection probably led to some neat image sharpening filters. But that’s more incidental than a direct link. Studying image data is a common ground between generating graphics and analyzing them. Researching one would’ve probably inevitably involved a little of the other. But the fundamentally statistical nature of machine learning/computer vision wasn’t too useful for computer graphics. There were experiments with Monte Carlo methods to cheapen ray-tracing, but the vast majority of rendering was done through plain triangle rasterization with some fancy effects. Image generation like the Electric Sheep project might’ve come close to merging the two, but I don’t know that any practical or commonly used filters came out of that experiment.
I guess image compression like arithmetic coding is pretty similar to ML since it’s basically “figure out the distribution of image data for all images and use it to directly encode the most common patterns in the fewest bits”, but that’s less image authorship and more image compression.
Background Pony #5108
I’ve had a few shower-thoughts I’d like to share.
- Let me offer counterexamples to the mistaken idea that modern “AI” image generation is utterly beholden to the images its data is based on, unable to produce anything new and just copy/pasting, collaging, blending, warping, and otherwise performing micro-plagiarism.
It’s hard to prove that something is original and not based on anything else. (In theory, with AI, it could be proved: someone could sift through the half-billion images in the source data and compare them to every tiny detail in the output. Good luck, though.) So, here’s a galaxy-brain Uno reverse card:
The image generator can be set up to closely replicate a brand-new photograph that was not in the data set - not re-training the AI, just using the new image as the goal and still requiring it to function with only the original data. It sounds backwards, arguing that copying something can prove it’s capable of making original content, but how else can you view the ability to produce something that’s provably not in the source data? That, or it is has so many tiny pieces to rearrange and blend to the point of pixel-level unrecognizability that it doesn’t even matter.
Consider also: AI systems can wind up copying things from the source data. If you type in “Van Gogh”, there’s a good chance you’ll see Starry Night because that’s so heavily represented. This is called “overfitting” and is considered a flaw to be engineered away from. The ability to not do that is an important part of why modern AI image generation is really heating up.
If you find people doing custom training for AIs and failing to avoid “overfitting” or even purposely letting it rip off poses, body shapes, faces, etc. absolutely call those specific people out for shitty behavior, just like you’d call out an illustrator who’s tracing. Don’t demonize AI image generation in general just like you wouldn’t demonize illustrators in general.
- I’m not convinced.
How about this:
A human can create something:
A human can create something:
- Completely new wholly imagined into a well-defined concept
- Something sufficiently well-described to them by someone else
- Something they’ve seen before
- Something influenced by any of the above, resulting in a new but derivative idea.
An AI can generate:
- Something they’ve seen before
- Something that’s a mix of things they’ve seen before
- Something that’s shaped by random noise and then interpreted into clear forms - so, fundamentally, new due to randomness instead of imagination, and refined by multiple millions of guidelines from the things they’ve seen.
A person could draw anything they’ve imagined, but to draw something specific like a pony, you’d need to learn about what a pony looks like. An AI also needs to know what a pony looks like.
To say “I don’t want my art used as AI training material” might as well also include, “I don’t want humans to look at my works out of fear that they’ll copy me, on purpose or subconsciously.” Perhaps you’d argue that a single specific human is less likely to copy, and that AI data collection gives powers to however many users to do with as they please, to which I’d like to remind you that too much similarity to the source data is a goal that’s actively avoided and to rightfully speak down to specific people who are negligently or even actively introducing “overfit” data that causes results to lean heavily on source imagery.
Throwing around terms like “art theft” just hurts what the term means when describing actual art theft such as tracing and plagiarism.
- It’s soulless. A machine can’t produce meaning.
Literally true, I guess. Good thing it’s humans that are consuming the images, free to read meaning into them based on what it looks like.
Oh! But it’s not true for the AI artists who do more than push a button and get an image - the ones who experiment, tweak, fix AI artifacts, rearrange the compositions, paint in new rough details, run it back through the AI software, etc. etc. Those could very well have soul.
- It takes no work. 5 minutes spent pasting in somebody’s prompt and changing a few words doesn’t make you an artist.
Just like a person might hold in very low regard an image made by barely recoloring a free base, or a scribble made in 5 minutes, or a matte painting that’s barely a few taps of a mega-hills-and-treeline brush preset - feel free to hold low-effort AI image-tool results in poor regard, and also celebrate the ones carefully crafted with an eye for artistry. There isn’t reason to condemn all AI image generation for art purposes and to dismiss the good, effortful results.
By the way, I get the sense you wouldn’t condemn a prodigy who can produce, at an age less than the number of years other artists have been practicing, wonderful paintings in 20m-1hr what other artists spend a day to make. I don’t think you’re really judging based on time and effort spent. That’s fine by me, but I’m confused why even pretend to care about that?
- It’s ugly. Anatomy is off. Limbs are weird.
Sounds like an issue of taste. But yeah, a lot of people are trying to use the tools to make believable humans, there’s definitely room for improvement. But like… That’s grounds for slamming the entire field as worthless? …have you SEEN art?
Exedrus
That’s a reasonable and optimistic prediction. It’s certainly in line with how past tech has changed art like that one background pony mentioned about photography. At some point, I’m sure we’ll see art programs advertising that users can paint some basic colors/tones into a layer, hit a button, and the program will convert those colors into trees/sky/dirt/lakes/etc. I think Nvidia did a demo for something like that a while ago.
I guess it boils down to just how far this tech goes. It’s possible that getting to the current state was relatively easy compared to getting these algorithms to respond sanely to the sorts of basic requests that artists can easily handle (i.e. “make her look more interested in X”, “I want her to be happier about Y”, “make Z look more like a mistake”). At that point artists wouldn’t have much to fear, just being able to talk with commissioners would be a huge advantage. That’s not what I’d bet on, but it’s really anyone’s guess what happens next.
Ciaran
(link is to one of the final vids)
Senior Moderator
友情は魔法だ
@Exedrus
This discussion of course is centered around ‘fine art’, but just looking at what’s happened with things like Zoom backgrounds and advertisements, I think that we’re only going to see more and more AI in consumer products and in entertainment. With things like Zoom it’s still operating in a ‘safe’ space, just making people’s backgrounds look less crappy in real time and adding a hat or cat ears to their heads - stuff like that.
This discussion of course is centered around ‘fine art’, but just looking at what’s happened with things like Zoom backgrounds and advertisements, I think that we’re only going to see more and more AI in consumer products and in entertainment. With things like Zoom it’s still operating in a ‘safe’ space, just making people’s backgrounds look less crappy in real time and adding a hat or cat ears to their heads - stuff like that.
But at some point tools that are currently very difficult to get working right, and that require tons of background work - like MMD - are going to be more and more real time, and I believe will become how we present ourselves in business meetings.
I mean, hell - just look at what people are doing to themselves in Zoom today.
Even when that happens though, people will still paint. Just like we still ride horses and still read physical books.
But just like horses are no longer dying in the streets of New York and books are becoming treasures to be cherished and displayed, as AI raises the bar for what art is common and ‘normal’ I think we’ll probably have fewer artists dying at their easels making minimum wage and working 20 hour days just to get out tomorrow’s TV ads.
Naughtypony2010 
Hoofsies Rule
I would like to bring some different vibes here.
Over the past couple days i have been thinking a lot and watching AI images and stuff. And I honestly now believe, people will always keep supporting real artists nonetheless. AI generated images, unless an obvious trace or currently rule breaking image from AI generating, should stay welcomed on derpibooru. I don’t see using countless reference works from pony art on this site being a theft.
Fellow artists, I will never support tracing or the pure intention of the loud idiot minority to replace artists who work their ass off to create a huge part of this fandom, make a living, bring joy to all of us.
AI generated pony images can result in some interesting things, a niche.
Something i have mentioned before, I might even look into it myself, perhaps get an even better understanding how it all works and see how difficult it might be to get a good result after all, right now and for years to come. Abstract and quite grotesque images already interest me quite a bit. things i have never seen from artists before. thats what i personally enjoy in AI images.
Something i have mentioned before, I might even look into it myself, perhaps get an even better understanding how it all works and see how difficult it might be to get a good result after all, right now and for years to come. Abstract and quite grotesque images already interest me quite a bit. things i have never seen from artists before. thats what i personally enjoy in AI images.
I’m not hacked I’m not turning over to the ‘dark side’ and again, will never support any kind of tracing, art theft, preference over artists who do what we all loved for a decade.
Love you all. Over and out, Naughtypony2010 <3 <3 <3 <3 <3
DerpyFast
I’m going to copy and paste my thoughts from the other thread:
As an artist, you don’t control how your art is interpreted, and you certainly can’t control who consumes your art. This fandom should know that better than anyone. Once you put something on the internet, you’re no longer truly in control of it.
There’s no way to affirmatively grant or deny permission for every possible thing a person could do. The builders of The Pyramids of Giza couldn’t grant photographers permission. Shakespeare can’t grant permission to have his works adapted into movies. Musicians from the 20s-40s couldn’t grant The Caretaker permission to do the editing that resulted in Everywhere At The End of Time.
As for AI itself, I watched an interesting documentary about AI that learned to play Go. An important thing to understand about AI is that it isn’t programmed, it’s bred. Where early Chess AI was just able to brute-force Chess, Go is much more complicated. Top Go players often have a hard time explaining exactly why a move is good. Not only can Go AI beat the top players in the world, it can win with moves that human players don’t even understand. This means that the AI is able to interpret information and draw its own conclusions.
Speaking of Chess AI, it’s gotten so powerful that humans can no longer beat the strongest ones. But the world of chess is thriving. One side effect of Chess AI is that it effectively gives us a way to objectively analyze chess games. When I was super into chess, I would record the moves, then plug them into the engine on chess.com to see what I did right, and what I could have done better.
As far as art goes, just because an AI could someday understand and execute the principles of art better than any human, that doesn’t necessarily translate to an easy way of telling the AI what to draw. As someone who’s done programming, I can tell you that it is orders of magnitude easier to tell a human what you want than it is to tell a machine. I think the real potential lies in an AI that “auto-completes” a rough sketch, or can non-verbally show the principles at work in a piece. I’ve got a ton of ideas that I simply lack the skill to realize. I think realizing the idea can be just as valid as the craft itself.
Interested in advertising on Derpibooru? Click here for information!
Help fund the $15 daily operational cost of Derpibooru - support us financially!