ChatGPT trains on its own output by exploring its own sampling space using reinforcement learning, Dalle 3 was trained on a dataset of synthetic captions, they are both SOTA, you can use AI to train AI models.
IT can be, though. High quality synthetic data for training is a thing, see Phi-2 [0] (and many more examples, people having been using GPT-4 outputs for finetuning a LOT to great avail).
> AI retraining on its own output is not a great thing.
For most forms of GenAI, including image generators, using existing models to generate synthetic data (and then doing human labelling, etc., as appropriate) for fine tuning is quite common and important. Why do you think its not a great thing?
Moreover, anything created with StabilityAI's code in its default settings also includes an invisible watermark in the image data, not just the generation information in the metadata.
> there is going to be a "pre-GPT" internet training set from 2022
Well, yeah, there are several here, and I think all the major image generators are using some combination of them as their starting points: https://laion.ai/
> As AI increases as an overall % of all online posts and activity it will death spiral on model quality.
Nope, it will just mean that it will be more expensive to source additional training data on top of the massive trove of existing "clean" (from intentional poisoning) data (much of which isn't perfectly captioned and human work on improving captioning can improve its utility in model training, as can more advanced models with more advanced text encoders, etc.)
If poisoning was widespread, it wouldn't impact "big model" quality much -- they aren't grabbing new random data on the internet for continuous training. It might drive up the expense of community fine tuning, which often does depend on sourcing representative imagery for target styles or concepts from, among other places, the internet.
This is just a race to the bottom and creating artificial work - folks creating poisoned data sets and then armies of data engineers who look at creating valid data sets for building models that can be grounded.
I feel for artists to an extent but no way are they going to succeed „ruining AI“ by poisoning image data sets.
And even if that was feasible, what the hell do they gain?
This is the whole „photography is not art“ or „digital image manipulation is not art“ or „oil painting is not art“ discussion all over again and I can believe serious artists actually take pet in this luddite theatre.
The lack of humanism in comments like this one are worrying. The fact that you don’t understand art or the artistic process doesn’t mean they are not relevant to a healthy society.
Plus it ignores the fact that all of our jobs are going to go away to automation and AI. Which can be a utopia other than we humans seem to be hell-bent on creating dystopias.
Well, i think it's strange to assume a serious artist would just take the first crop given to them and not alter at all! After all, there are numerous ai tools for altering ai images to fit your choices better. You simply pick the best template and then direct the AI, with intent, at exactly which parts of the image you'd like changed and into what. It's a delicate balance of prompt engineering (knowing exactly what to say, sometimes terms not know by the lay public become necessary for the best creation of your idea), and actually having an artistic idea in mind. It takes inspiration, tact, composition, knowledge, etc.
Sure, putting in "throwing up technicolor on the burger king riding a giant snail" is incredibly easy. Being able to fix everything the AI cannot interpret about your ideal, let alone having an ideal strong enough to differ from what your presented is artistic. I'm well aware there are people vehemently opposed to this idea, but there should be an advocate for the opposition
I concur, and I don't understand why a lot of artists still talk about AI art as if prompt -> end result is the only way of interacting with these systems. That may have been true in Spring of last year, but entire pipelines, flows, tool integrations, models etc. have been created since then. To really excert control over the generative process, you need more than Dall-E through ChatGPT, you need ControlNet, you need various models, and so on. It's technical artistry as 3D modeling is technical artistry.
> Putting some prompts into an image generator and choosing which one you like is hardly the same thing.
At a base level, "putting some prompts [..] and choosing" sounds like intent to me. A user is choosing what they want to capture and making a choice, much like I, as a photographer, look for a particular scene (or happen across one!) and choose how I want to capture it.
Of course, you can do much more in terms of control of image synthesis these days. ControlNet and friends make it feasible to control every aspect of the resulting image.
> A user is choosing what they want to capture and making a choice, much like I, as a photographer, look for a particular scene (or happen across one!) and choose how I want to capture it.
The only similarities between these is the fact that they’re making a choice. The act of making a choice does very little to indicate that the underlying process are similar enough to be compared to each other.
The choice made by the photographer is made based on a myriad of factors and intuitions. The result is a snapshot of a real moment in time, modulated by the photographer’s artistic ability, technical prowess and unique perspective of the world based on a lifetime of experience.
Choosing what I make for breakfast is also a form of choosing.
Why do those things not apply to someone using AI? I hate making this kind of argument, but:
The choice made by the AI artist is made based on a myriad of factors and intuitions. The result is a snapshot of the latent space, modulated by the artist's artistic girl [grit?], technical prowess and unique perspective of the world based on a lifetime of experience.
You still need those things in order to make a compelling image with AI, especially if you're getting into the more advanced uses.
> Why do those things not apply to someone using AI? I hate making this kind of argument, but:
I'm glad you made this argument, because I think it helps illustrate the point. When we talk about photography vs. AI vs. other forms of artwork in highly abstract terms, they all sound superficially similar.
The moment you start to dive into the details behind each process, it's clear that they are entirely unlike each other. So yes, the AI "artist" is drawing on a myriad of factors and intuitions, but the nature of those factors and intuitions and the effect they have on the output are drastically different to the point that they don't resemble photography even a little bit.
For example, the photographer who wants to take a stunning photograph of a snowy mountain vista at sunrise often has to:
- Use their knowledge of basic astronomy and sun movement to plan the timing and location of the photo
- Use their body to physically hike to the location that will give them this stunning view
- Use their knowledge of survival to stay safe while they hike
- Once on location, use their knowledge of perspective, depth, camera angle, and composition to find a pleasing place to take the photo from (composition is one of the single most impactful choices and only comes from practice/skill)
- While under time pressure from the rising sun, dial in the correct depth of field and exposure
- When things go wrong (which they often do), quickly re-assess all of the above and try shots from new location
- When done, understand the properties of light and color to a degree that enables them to develop the photo so that it matches the original scene
- The wildlife photographer draws on yet another deep and potentially vast skillset
The AI "artist" types "a stunning photograph of a snowy mountain vista at sunrise" into a box and clicks the Generate button.
It may be that knowing how to formulate a sentence that describes the desired scene relies on a myriad of factors and intuitions, but to compare those factors and intuitions to those the photographer is using seems deeply problematic.
Which leads us to reason people find stunning photographs enjoyable to look at in the first place. Encoded in that photograph is the story of the person who took it. In every piece of traditional art (using the word traditional here to mean not AI), the artist's life experiences, thoughts, emotions, and personal investment are baked into the result. This is why many/most people find art compelling. Not just for its beauty, but for the humanity encoded into it.
AI imagery is its own form of deeply fascinating when you think of it as a distillation of the skills of millions of people baked into a single model. But when you see a piece of AI imagery, it cannot evoke the same kind of emotional connection to the source, because it's far removed from the original artistic process. It's something, but it's not compelling.
> You still need those things in order to make a compelling image with AI, especially if you're getting into the more advanced uses.
You need very little, relatively speaking, to make a compelling image with AI when compared to the life experiences of the people who made the imagery that powers current AI models. That's not to say you need nothing, but I don't think they can be compared.
Once you move beyond these abstract comparisons, the underlying processes show themselves to be entirely unlike each other.
I should also mention that I'm not saying AI imagery has no place. I think it's tremendously useful for the creation of educational content, to help people communicate ideas they previously didn't have the skills to communicate, etc. But I think it sits in a very different category.
One could say much the same comparing photography to painting. Photography allows people to communicate in ways that previously they didn't have the skill to with paint. AI is just a tool, the art will be in learning how to best use it.
One could say many things about the difference between photography and painting of course, but I don't think the things said about these differences are "much the same". In the abstract, they appear superficially similar. When you consider the underlying details, they are worlds apart, and I think it's a category error to compare them directly.
Painting is painting. Photography is photography. AI images are AI images (a category that will be further subdivided as the technology matures, much like painting and photography make up a spectrum of skills and approaches).
They all sit in their own spaces, each with characteristics that make them more or less interesting. I'm personally far more interested in the kind of art that requires investment of human capital: mental, emotional, talent, skill, time, effort. I care about quality paintings and photos not just because of what they look like, but because of what it took to create them. Because of the emotions intrinsic to every brush stroke. Or the immersion of the photographer's consciousness in a real location that was difficult to reach. That human investment is core to the expression of the work, and to the interest it garners.
I do think AI tools will be transformative for traditional artists and will unlock new possibilities. They'll also enable a new generation of artists to turn their imaginations into visual artifacts they were previously not capable of creating. I think the same is true for LLMs and writing. But at the core of it all, these are all forms of human-to-human communication, and the process itself carries much of the message. Much of that message is lost to the medium when we abstract more and more of the process away.
As a thought experiment, consider food replicators in Star Trek. I can tell it what I want, and food appears. This is entirely unlike the process of making a high quality meal from scratch, and I wouldn't call myself a chef for having the ability to describe to a machine what I'd like it to do. This isn't to say there is no value in knowing how to instruct the machine well, but it's just a very different category of human activity.
Art is art. The argument is just a pile of shifting sands. But artists make choices! But artists require knowledge! But artists need to communicate!
A good artist will be able to use AI tools to create compelling art that communicates through decisive choices of their design. Other artists will get lucky and let the computer do most of the work.
Art is a word that humans made up to categorize a form of human expression.
Whatever art is, it is because of the value that humans have chosen to ascribe to it, which they do for countless reasons. Many of those reasons are simply not present in AI output, and AI will bring new reasons that will be compelling for some, but entirely uninteresting for others.
Clinging to an abstract three letter word as the ultimate definition of something nearly undefinable ignores both the richness of the form and the sociological factors that make the word hold any meaning.
Again, I’m not saying that artists won’t do great things with AI. I’m saying that the work done with AI will not hold the same kind of value that is intrinsic to other forms. Some people won’t care about the underlying process and it sounds like you may be in that category. Some people will continue to care, and I’d count myself in that category.
Many people point out that Photography is to Painting what AI is to Photography. And while I think the analogy is a bit stretched, the primary reason this comparison comes up is because photography is thought of as a “lesser skilled” method of producing images. And I think that’s a fair assessment. It is not debatable that producing a photorealistic or highly skilled representation of a scene with a paintbrush requires far more skill and far more effort than operating a camera, especially in the 2020s. If your enjoyment of a thing comes from appreciating a particular kind of process, e.g. you are fascinated by and enjoy the output of skilled painters, not just for the result, but for the entire process, then AI (or Photography) isn’t going to hold your interest.
Coffee is coffee. But there sure is a large spectrum of possible ways to experience coffee. Many people are happy with the Folgers instant stuff. Connoisseurs have a palate that knows that there’s more to coffee than caffeine and brown water, but not everyone is a connoisseur and some people are more interested in the utility of coffee (caffeine/energy).
That isn’t to say that instant coffee is not coffee, and people are welcome to enjoy it if they can. But it sure isn’t the same as a cup of recently roasted, freshly ground single origin beans carefully brewed at a temperature calibrated to the properties of the bean and roast. Foljers will wake me up. That carefully brewed cup will wake me up and produce a highly pleasurable taste experience that can alter my emotional state in a positive direction.
My point is that nuance abounds, and trying to reduce something like art of all things to a one dimensional description seems not very useful.
> work done with AI will not hold the same kind of value that is intrinsic to other form
It will hold a different kind of value. What that means to the individual is something they have to decide for themselves.
I for one love new things and progress, but I can also understand a more conservative mindset even if I feel differently.
I am also tired of people screeching „AI is not art!!1!“ and trying to deplatform and attack people who feel otherwise (not insinuating that you do this by the way).
Yeah, I agree it will hold a different kind of value.
I don’t think people should be screeching about this either. I do think people are screeching because they enjoy traditional art and they think traditional art is threatened.
I’m starting to see this more like computerized chess. When a computer first beat a human, it didn’t end chess. Chess is alive and well, because humans like playing chess. With that said, the chess community is also more susceptible to cheating than it has ever been, and has suffered controversy as a result.
I think AI art will be similar. People who want traditional art will keep wanting it. People who create it will keep creating it. The people on both sides of this exchange aren’t going to suddenly stop valuing the things they valued before.
But like chess, AI tools make it easy for people to misrepresent themselves, and people will absolutely try to pass off their creations as original. I suspect that over time, tools will catch up and people who eschew AI tooling will be valued for it. It’s just gonna be a mess for awhile.
> I’m saying that the work done with AI will not hold the same kind of value that is intrinsic to other forms.
The value is literally not intrinsic, it is extrinsic and specifically socially/culturally ascribed.
And, yes, that's also true of different forms of art; while all have value, the kind and degree of value in any particular society of sculpture vs. representational oil painting vs. abstract oil painting vs. photography vs. 3d-rendered digital art are all different (and the difference between them aren't constant between times and places where they are present, either.)
Yes, art produceed using, e.g., diffusion models will be different from all the others in this regard, but different in the same way that all the others already differ among themselves.
Intrinsic was a poor word choice on my part. My point was that some of the most valued aspects of most art up to this point - the investment of human talent/skill and the uniqueness brought by the individual - is either not present, or is hard to define when dealing with AI art.
> but different in the same way that all the others already differ among themselves
I’d argue it’s different in a very different way. The conceptual gap between AI and traditional forms of art is arguably much wider than the gaps/differences that already exist between traditional forms.
Using mental models to reason about AI art that liken it to those older forms can be misleading. I do agree that it’s a new form that will carry its own properties and to which people will ascribe different kinds of value. Whether or not that ascribed value is deserved is another question - a question that must be raised because again, someone with artistic skills was still required to create the AI output, they’re just long forgotten by the time the image is generated.
If nothing else, the massive unanswered ethical questions raised by the use of models that can only exist because of the content created by traditional artists puts this firmly in its own category.
And of course, for every user that used to be an artist before AI, there are now 10,000 such users wielding their new-found artistic intent via prompting.
There's a lot more to AI art generation than "putting some prompts into an image generator". Sure, putting some prompts into a generator _also_ counts as "art" in the same manner as "child scribbles doodles on paper" counts as art. But you're dismissing an entire field of aspiring artists without having even a basic understanding of what AI artists do to achieve the images they publish. Go watch some videos of decent AI art workflows and then come back.
If you believe you can’t use Stable Diffusion for example with intent, I don’t believe you ever seriously used it.
There is a ton of work going into it, from training your own loras, building workflows that achieve what you are looking for, using img2img and literally dozens of other related tools and methods.
With the same argument you can say photography doesn’t have intent - you just point the camera somewhere and press the button.
Edit: for what it’s worth, Duchamp just stole the toilet from another artist and passed it off as his own work - I would argue that is much worse than using something like SD.
There is a whole lot of art that is not "high art". Just commerical art. Graphic design. Not all of it is "paintings in hotel rooms"; it's all over the place. People have jobs in this area.
>I can believe serious artists actually take pet in this luddite theatre.
I think more people should participate in the luddite theatre. Luddites weren't against automation and technological advances per se but that the profits of increased productivity weren't shared. The people that are profiting off AI generated art today don't want to participate in distributing their profits more widely, despite generative models being based on the works of other people, some of whose art was used for the training without permission.
I’m tired of the “they’re Luddites” middlebrow dismissal, ignoring all of the history and lessons of that time.
Yeah, Industrialization was a completely painless and smooth process. No one was ever inconvenienced. That’s why we can make fun of the then-naysayers now.
That is fair.
I mostly wanted to express that it’s not going to accomplish their goal. Image generators are here to stay no matter how hard and long you fight against them.
Of course technological progress is never a painless process, it’s in fact extremely violent and brutal. On the other hand, industrialisation made it possible to feed 10 billion humans and eradicated much of the everyday pain humans had to deal with since the Stone Age. This would not have been possible otherwise.
> That is fair. I mostly wanted to express that it’s not going to accomplish their goal. Image generators are here to stay no matter how hard and long you fight against them.
What’s fair? You’re just repeating what you already said.
Image generators are here to stay? No kidding. Kind of like the machines of Industrial Revolution? Yeah, obviously. And people fought back when they in turn started to be treated like cogs in the Machine. And a lot of that worked. Even worked so successfully that we can dismiss those old fogies as backwards and simple. So privileged are we that we don’t have to know the history of what they did for us.
> Of course technological progress is never a painless process, it’s in fact extremely violent and brutal. On the other hand, industrialisation made it possible to feed 10 billion humans and eradicated much of the everyday pain humans had to deal with since the Stone Age. This would not have been possible otherwise.
What on the other hand? Why do you think we have the quality of life that we do? Because people sat on their hands and trusted the invisible hand of progress to sort things out for them? No, it is exactly because they fought back that things are working so well for us today.
But whatever. I’ve got my microwave oven and my fridge. Suck on that, stupid Luddites.
Hardly comparable. The fact that people can generate work doesn't mean they're owed labor because their work is being automated away. The physics of work is not a creative endeavor and even though I think ownership of creative works is grey matter the idea of having to share profits after generating more work with machines is just blindingly illogical
But why do they have to participate in the profits? How musicians participate the profits of digital audio workstations that allow people to make music without real musicians? How did older artists profit from current artists that learned, and oftentimes virtually copied their style? The solution for artists is to embrace these tools to make even better art.
I remember when Waze and Google Maps came out and angry neighborhood residents would try to "poison" the traffic data set by reporting non-existant police in their neighborhood streets.
Yeah, this is really just artists poisoning themselves hoping AI dies. It'll take away from them using their creativity to actually capitalize and they'll just spend their time trying to go against relatively intelligent people on their own turf.
> And even if that was feasible, what the hell do they gain?
More visibility by the world through news articles exactly like this?
More visibility perhaps clues more average joes of the world into what exactly is going on. Maybe it gains them more people who might be on their side and have sympathy for them, etc.
Even if they don't have a clear end game, doing something, anything they can probably feels better to them than doing nothing, and just going quietly into the night without putting up any resistance.
This is not the same as those things at all. The poisoning of the data sets only works because the AI is scraping the internet for images and stealing the artist's works. They are mad because their art is being stolen and they didn't consent to it.
It's easy to call other people names when it's not your life's work at stake.
A nanoscopic amount of artists sabotaging a few images will not do anything to anyone, there'll just be a small version bump in vision models and the 0.0000001 contaminated images will also be classified correctly.
While i really feel for the artist like everyone else potentially losing their livelihood to AI, this is an annoying fantasy story.
Just wait for luddite political candidates becoming popular when white collar workers will start be inevitable replaced. It will be a total disaster for any sensible discourse.
I think the US already went through their populist cycle between 2008-2020. If that cycle theory is true, we shouldn't have to worry about it until our old days.
I dislike the unquestioned assumption that "Luddite bad, machine good".
If you love progress for progress' sake, think about where you're progressing _to_. So far we've "progressed":
- My furniture no longer being made by skilled carpenters: Everything is falls apart after a couple of years, and great, I have to put it together myself to begin with
- Lowering of the bar to _literally anyone with a computer_ for music creation: I know all old people say this, so I won't break tradition: Your music is shite
- Hi-Fi not being particularly hi-fi: A Bluetooth receiver playing lossy format crap out of terrible speakers
- Cashiers to automatic checkouts: Cool, now I have to do their job without getting
paid, then wait for the one staff member they kept on payroll to come and check my shopping and rescan everything because of course they don't trust me
I could go on all night. Point is; progress just means forward motion, not always in a good direction.
Can you explain how extinguishing the artists or, its practical equivalent, destroying the economic incentives for them to exist is being in the right?
Poison works by being able to scale its impact dramatically: attacking chemical cornerstone processes, disabling a receptor, etc.
ML is, to simplify, a statistical process that tries to identify probability distributions. That implies they’d have to purposefully degrade large amounts of images that the system has access to. Unless those images are impossible to distinguish from the others (because, say, they come from the same sites) I’m not sure how they’d do it consistently.
They could mark sources, by using an unused combination of keywords and having a unique image attached to it, but poisoning—at scale? I’m not seeing a stable equilibrium where that’s possible.
I agree with people this is a long shot, but how can you not admire the spunk? Whether we realize it or not yet, this is the kind of resolve and energy we need to bring to the future.
Antagonism is not automatically luddite-ism! This is capitalism right? Dont we need the competition?
We are on a roller coast ride of technological advancement. Gravity is pulling us inexorably forward, and we can't see where the track eventually goes. We have brakes, but not enough to stop forward movement entirely. We have some input on the direction we take, but our limited vision down the track makes it hard for us to anticipate and correct for any upcoming curves we don't like.
It's going to be a bumpy ride, but hopefully we have fun along the way.
I think you misunderstand the violence that lies in our near future. Before, when we began automating manual labor we proclaimed you should get educated and design the machines and improve the processes. Now, we're developing tools to displace large swaths of those workers. There simply aren't other jobs for them to do that we haven't already automated. History shows this is not sustainable and will lead to protest and violence. I would suggest we avoid that, after all, that would be the smart thing to do, but our species has proved time and time again that it's not smart in the aggregate.
It's never going to matter, ignoring that they are a extremely small percentage of images on internet, if they have enough signal to almost not be noticed by a human than there is more than enough signal to train an AI.
> altering an image’s pixels in a way that wreaks havoc to computer vision but leaves the image unaltered to a human’s eyes
Can someone who understands the original paper give an ELI5 on how that's possible?
I understand how labeling an image of a cat as "dog" could poison a dataset.
I also understand how adding images of a toaster (which sort of looks like a handbag) to a dataset of handbags could poison a dataset.
But I don't understand what's happening in Figure 6 in the original paper. The pairs of pictures of the dogs, cars, etc. look absolutely identical. What exactly is happening there?
Computer vision systems are not as robust as human vision systems. There are subtle changes to pixel values that can utterly flummox a computer vision system, but seem undetectable to a human looking at the image.
It varies. Specially crafted noise added to the pixel values. Looks like random noise, but obviously isn't. TBH I'm not an expert, but as I understand it, It is "trained" using the vision network, with a loss function that is some combination of being low amplitude, and reducing the strength of the correct image identification.
This is a guess (I have read some of the paper but to me it does not seem to explain), but presumably you work backwards from the result you want. There is some vector that comes out of the classifier model that would represent cat, and another that would represent dog; you get the difference between these two models, and work back through the layers of matrix multiplication finding places where small perturbations in the input data force the classifier to make the exact errors you need.
They claim that the poisoning is transferrable across different models, which would not be the case if that was true.
OTOH, given this group's track record with their earlier poisoning effort (Glaze) vs. the adulatory press that they managed to arrange, I don't expect much from Nightshade despite the similarly adulatory press. Great media relations game, though.
Artists are happy to use the global working plattforms (Instagram, ...) to show and monetize their art to a world-wide audience but are unhappy if these content is used to train AI. If you show your art in just your own local gallery nobody will exploit it. Its the same in software engineering with public repos. So I guess we all have to adjust and use AI for the better instead of complaining and fighting it. Or am I missing the point here?
This reminds me of an attempt when spammers tried (and still trying, but not as often as before) to poison antispam bayes filters with all sorts of garbage in email. The result is that filters became even more powerful over time.
"What if I take my own pictures, and train the data on my own pictures?"
Their answer:
"No"
Luddites, nothing more. They will become obscure while everyone else advances. She uses her colored pencils, I use digital art. Guess who works at a coffee shop.
It reminds me of a wave of comments "Deadline tomorrow !!! Everything you’ve ever posted becomes public from tomorrow. I give notice to Facebook it is strictly forbidden to disclose, copy, distribute, or take any other action against me. If you do not publish a statement at least once it will be tactically allowing the use of your photos, as well as the information contained in the profile status updates. DO NOT SHARE. Copy and paste."
But now people are posting images instead of text. And it all looks so stupid that it only deals reputational blow to some subgroup of artists who now look like a group of technically illiterate luddites.
103 comments
[ 4.5 ms ] story [ 195 ms ] threadAs AI increases as an overall % of all online posts and activity it will death spiral on model quality.
AI retraining on its own output is not a great thing.
[0] https://www.microsoft.com/en-us/research/blog/phi-2-the-surp...
For most forms of GenAI, including image generators, using existing models to generate synthetic data (and then doing human labelling, etc., as appropriate) for fine tuning is quite common and important. Why do you think its not a great thing?
I see no reason to believe there will be a death spiral on model quality - in the last year or so the opposite has been happening.
Of course the jury is still out on the future.
Do they now?
Well, yeah, there are several here, and I think all the major image generators are using some combination of them as their starting points: https://laion.ai/
> As AI increases as an overall % of all online posts and activity it will death spiral on model quality.
Nope, it will just mean that it will be more expensive to source additional training data on top of the massive trove of existing "clean" (from intentional poisoning) data (much of which isn't perfectly captioned and human work on improving captioning can improve its utility in model training, as can more advanced models with more advanced text encoders, etc.)
If poisoning was widespread, it wouldn't impact "big model" quality much -- they aren't grabbing new random data on the internet for continuous training. It might drive up the expense of community fine tuning, which often does depend on sourcing representative imagery for target styles or concepts from, among other places, the internet.
This is the whole „photography is not art“ or „digital image manipulation is not art“ or „oil painting is not art“ discussion all over again and I can believe serious artists actually take pet in this luddite theatre.
a few more years of employment, at best. It sucks, but it's reality. There's no stopping the onslaught of AI.
Better start learning how to take advantage of the AI tools to make yourself not irrelevant.
if they are relevant, then the proof would be that they'd be paid to create art by other humans who believe this notion of relevency.
Many less years than if they learned the new tools that magnify productivity.
Putting some prompts into an image generator and choosing which one you like is hardly the same thing.
However, the idea of AI generated images could be considered art (a bit like Duchamp's toilet), but it is pretty much done now.
Of course, 'paintings for hotel rooms' is something different, and indeed may be affected by AI
Sure, putting in "throwing up technicolor on the burger king riding a giant snail" is incredibly easy. Being able to fix everything the AI cannot interpret about your ideal, let alone having an ideal strong enough to differ from what your presented is artistic. I'm well aware there are people vehemently opposed to this idea, but there should be an advocate for the opposition
At a base level, "putting some prompts [..] and choosing" sounds like intent to me. A user is choosing what they want to capture and making a choice, much like I, as a photographer, look for a particular scene (or happen across one!) and choose how I want to capture it.
Of course, you can do much more in terms of control of image synthesis these days. ControlNet and friends make it feasible to control every aspect of the resulting image.
The only similarities between these is the fact that they’re making a choice. The act of making a choice does very little to indicate that the underlying process are similar enough to be compared to each other.
The choice made by the photographer is made based on a myriad of factors and intuitions. The result is a snapshot of a real moment in time, modulated by the photographer’s artistic ability, technical prowess and unique perspective of the world based on a lifetime of experience.
Choosing what I make for breakfast is also a form of choosing.
The choice made by the AI artist is made based on a myriad of factors and intuitions. The result is a snapshot of the latent space, modulated by the artist's artistic girl [grit?], technical prowess and unique perspective of the world based on a lifetime of experience.
You still need those things in order to make a compelling image with AI, especially if you're getting into the more advanced uses.
I'm glad you made this argument, because I think it helps illustrate the point. When we talk about photography vs. AI vs. other forms of artwork in highly abstract terms, they all sound superficially similar.
The moment you start to dive into the details behind each process, it's clear that they are entirely unlike each other. So yes, the AI "artist" is drawing on a myriad of factors and intuitions, but the nature of those factors and intuitions and the effect they have on the output are drastically different to the point that they don't resemble photography even a little bit.
For example, the photographer who wants to take a stunning photograph of a snowy mountain vista at sunrise often has to:
- Use their knowledge of basic astronomy and sun movement to plan the timing and location of the photo
- Use their body to physically hike to the location that will give them this stunning view
- Use their knowledge of survival to stay safe while they hike
- Once on location, use their knowledge of perspective, depth, camera angle, and composition to find a pleasing place to take the photo from (composition is one of the single most impactful choices and only comes from practice/skill)
- While under time pressure from the rising sun, dial in the correct depth of field and exposure
- When things go wrong (which they often do), quickly re-assess all of the above and try shots from new location
- When done, understand the properties of light and color to a degree that enables them to develop the photo so that it matches the original scene
- The wildlife photographer draws on yet another deep and potentially vast skillset
The AI "artist" types "a stunning photograph of a snowy mountain vista at sunrise" into a box and clicks the Generate button.
It may be that knowing how to formulate a sentence that describes the desired scene relies on a myriad of factors and intuitions, but to compare those factors and intuitions to those the photographer is using seems deeply problematic.
Which leads us to reason people find stunning photographs enjoyable to look at in the first place. Encoded in that photograph is the story of the person who took it. In every piece of traditional art (using the word traditional here to mean not AI), the artist's life experiences, thoughts, emotions, and personal investment are baked into the result. This is why many/most people find art compelling. Not just for its beauty, but for the humanity encoded into it.
AI imagery is its own form of deeply fascinating when you think of it as a distillation of the skills of millions of people baked into a single model. But when you see a piece of AI imagery, it cannot evoke the same kind of emotional connection to the source, because it's far removed from the original artistic process. It's something, but it's not compelling.
> You still need those things in order to make a compelling image with AI, especially if you're getting into the more advanced uses.
You need very little, relatively speaking, to make a compelling image with AI when compared to the life experiences of the people who made the imagery that powers current AI models. That's not to say you need nothing, but I don't think they can be compared.
Once you move beyond these abstract comparisons, the underlying processes show themselves to be entirely unlike each other.
I should also mention that I'm not saying AI imagery has no place. I think it's tremendously useful for the creation of educational content, to help people communicate ideas they previously didn't have the skills to communicate, etc. But I think it sits in a very different category.
Painting is painting. Photography is photography. AI images are AI images (a category that will be further subdivided as the technology matures, much like painting and photography make up a spectrum of skills and approaches).
They all sit in their own spaces, each with characteristics that make them more or less interesting. I'm personally far more interested in the kind of art that requires investment of human capital: mental, emotional, talent, skill, time, effort. I care about quality paintings and photos not just because of what they look like, but because of what it took to create them. Because of the emotions intrinsic to every brush stroke. Or the immersion of the photographer's consciousness in a real location that was difficult to reach. That human investment is core to the expression of the work, and to the interest it garners.
I do think AI tools will be transformative for traditional artists and will unlock new possibilities. They'll also enable a new generation of artists to turn their imaginations into visual artifacts they were previously not capable of creating. I think the same is true for LLMs and writing. But at the core of it all, these are all forms of human-to-human communication, and the process itself carries much of the message. Much of that message is lost to the medium when we abstract more and more of the process away.
As a thought experiment, consider food replicators in Star Trek. I can tell it what I want, and food appears. This is entirely unlike the process of making a high quality meal from scratch, and I wouldn't call myself a chef for having the ability to describe to a machine what I'd like it to do. This isn't to say there is no value in knowing how to instruct the machine well, but it's just a very different category of human activity.
A good artist will be able to use AI tools to create compelling art that communicates through decisive choices of their design. Other artists will get lucky and let the computer do most of the work.
Whatever art is, it is because of the value that humans have chosen to ascribe to it, which they do for countless reasons. Many of those reasons are simply not present in AI output, and AI will bring new reasons that will be compelling for some, but entirely uninteresting for others.
Clinging to an abstract three letter word as the ultimate definition of something nearly undefinable ignores both the richness of the form and the sociological factors that make the word hold any meaning.
Again, I’m not saying that artists won’t do great things with AI. I’m saying that the work done with AI will not hold the same kind of value that is intrinsic to other forms. Some people won’t care about the underlying process and it sounds like you may be in that category. Some people will continue to care, and I’d count myself in that category.
Many people point out that Photography is to Painting what AI is to Photography. And while I think the analogy is a bit stretched, the primary reason this comparison comes up is because photography is thought of as a “lesser skilled” method of producing images. And I think that’s a fair assessment. It is not debatable that producing a photorealistic or highly skilled representation of a scene with a paintbrush requires far more skill and far more effort than operating a camera, especially in the 2020s. If your enjoyment of a thing comes from appreciating a particular kind of process, e.g. you are fascinated by and enjoy the output of skilled painters, not just for the result, but for the entire process, then AI (or Photography) isn’t going to hold your interest.
Coffee is coffee. But there sure is a large spectrum of possible ways to experience coffee. Many people are happy with the Folgers instant stuff. Connoisseurs have a palate that knows that there’s more to coffee than caffeine and brown water, but not everyone is a connoisseur and some people are more interested in the utility of coffee (caffeine/energy).
That isn’t to say that instant coffee is not coffee, and people are welcome to enjoy it if they can. But it sure isn’t the same as a cup of recently roasted, freshly ground single origin beans carefully brewed at a temperature calibrated to the properties of the bean and roast. Foljers will wake me up. That carefully brewed cup will wake me up and produce a highly pleasurable taste experience that can alter my emotional state in a positive direction.
My point is that nuance abounds, and trying to reduce something like art of all things to a one dimensional description seems not very useful.
It will hold a different kind of value. What that means to the individual is something they have to decide for themselves.
I for one love new things and progress, but I can also understand a more conservative mindset even if I feel differently.
I am also tired of people screeching „AI is not art!!1!“ and trying to deplatform and attack people who feel otherwise (not insinuating that you do this by the way).
I don’t think people should be screeching about this either. I do think people are screeching because they enjoy traditional art and they think traditional art is threatened.
I’m starting to see this more like computerized chess. When a computer first beat a human, it didn’t end chess. Chess is alive and well, because humans like playing chess. With that said, the chess community is also more susceptible to cheating than it has ever been, and has suffered controversy as a result.
I think AI art will be similar. People who want traditional art will keep wanting it. People who create it will keep creating it. The people on both sides of this exchange aren’t going to suddenly stop valuing the things they valued before.
But like chess, AI tools make it easy for people to misrepresent themselves, and people will absolutely try to pass off their creations as original. I suspect that over time, tools will catch up and people who eschew AI tooling will be valued for it. It’s just gonna be a mess for awhile.
The value is literally not intrinsic, it is extrinsic and specifically socially/culturally ascribed.
And, yes, that's also true of different forms of art; while all have value, the kind and degree of value in any particular society of sculpture vs. representational oil painting vs. abstract oil painting vs. photography vs. 3d-rendered digital art are all different (and the difference between them aren't constant between times and places where they are present, either.)
Yes, art produceed using, e.g., diffusion models will be different from all the others in this regard, but different in the same way that all the others already differ among themselves.
> but different in the same way that all the others already differ among themselves
I’d argue it’s different in a very different way. The conceptual gap between AI and traditional forms of art is arguably much wider than the gaps/differences that already exist between traditional forms.
Using mental models to reason about AI art that liken it to those older forms can be misleading. I do agree that it’s a new form that will carry its own properties and to which people will ascribe different kinds of value. Whether or not that ascribed value is deserved is another question - a question that must be raised because again, someone with artistic skills was still required to create the AI output, they’re just long forgotten by the time the image is generated.
If nothing else, the massive unanswered ethical questions raised by the use of models that can only exist because of the content created by traditional artists puts this firmly in its own category.
With the same argument you can say photography doesn’t have intent - you just point the camera somewhere and press the button.
Edit: for what it’s worth, Duchamp just stole the toilet from another artist and passed it off as his own work - I would argue that is much worse than using something like SD.
I think more people should participate in the luddite theatre. Luddites weren't against automation and technological advances per se but that the profits of increased productivity weren't shared. The people that are profiting off AI generated art today don't want to participate in distributing their profits more widely, despite generative models being based on the works of other people, some of whose art was used for the training without permission.
Yeah, Industrialization was a completely painless and smooth process. No one was ever inconvenienced. That’s why we can make fun of the then-naysayers now.
Of course technological progress is never a painless process, it’s in fact extremely violent and brutal. On the other hand, industrialisation made it possible to feed 10 billion humans and eradicated much of the everyday pain humans had to deal with since the Stone Age. This would not have been possible otherwise.
What’s fair? You’re just repeating what you already said.
Image generators are here to stay? No kidding. Kind of like the machines of Industrial Revolution? Yeah, obviously. And people fought back when they in turn started to be treated like cogs in the Machine. And a lot of that worked. Even worked so successfully that we can dismiss those old fogies as backwards and simple. So privileged are we that we don’t have to know the history of what they did for us.
> Of course technological progress is never a painless process, it’s in fact extremely violent and brutal. On the other hand, industrialisation made it possible to feed 10 billion humans and eradicated much of the everyday pain humans had to deal with since the Stone Age. This would not have been possible otherwise.
What on the other hand? Why do you think we have the quality of life that we do? Because people sat on their hands and trusted the invisible hand of progress to sort things out for them? No, it is exactly because they fought back that things are working so well for us today.
But whatever. I’ve got my microwave oven and my fridge. Suck on that, stupid Luddites.
That also didn't work.
More visibility by the world through news articles exactly like this?
More visibility perhaps clues more average joes of the world into what exactly is going on. Maybe it gains them more people who might be on their side and have sympathy for them, etc.
Even if they don't have a clear end game, doing something, anything they can probably feels better to them than doing nothing, and just going quietly into the night without putting up any resistance.
It's easy to call other people names when it's not your life's work at stake.
While i really feel for the artist like everyone else potentially losing their livelihood to AI, this is an annoying fantasy story.
If you love progress for progress' sake, think about where you're progressing _to_. So far we've "progressed":
- My furniture no longer being made by skilled carpenters: Everything is falls apart after a couple of years, and great, I have to put it together myself to begin with
- Lowering of the bar to _literally anyone with a computer_ for music creation: I know all old people say this, so I won't break tradition: Your music is shite
- Hi-Fi not being particularly hi-fi: A Bluetooth receiver playing lossy format crap out of terrible speakers
- Cashiers to automatic checkouts: Cool, now I have to do their job without getting paid, then wait for the one staff member they kept on payroll to come and check my shopping and rescan everything because of course they don't trust me
I could go on all night. Point is; progress just means forward motion, not always in a good direction.
The people who are happy to have higher quality artwork at lower costs, allowing humanity to progress further?
Or the establishment who want their own desires fulfilled at the expense of the rest of humanity?
ML is, to simplify, a statistical process that tries to identify probability distributions. That implies they’d have to purposefully degrade large amounts of images that the system has access to. Unless those images are impossible to distinguish from the others (because, say, they come from the same sites) I’m not sure how they’d do it consistently.
They could mark sources, by using an unused combination of keywords and having a unique image attached to it, but poisoning—at scale? I’m not seeing a stable equilibrium where that’s possible.
https://news.ycombinator.com/item?id=38013151
https://news.ycombinator.com/item?id=37990750
Antagonism is not automatically luddite-ism! This is capitalism right? Dont we need the competition?
It's going to be a bumpy ride, but hopefully we have fun along the way.
Can someone who understands the original paper give an ELI5 on how that's possible?
I understand how labeling an image of a cat as "dog" could poison a dataset.
I also understand how adding images of a toaster (which sort of looks like a handbag) to a dataset of handbags could poison a dataset.
But I don't understand what's happening in Figure 6 in the original paper. The pairs of pictures of the dogs, cars, etc. look absolutely identical. What exactly is happening there?
OTOH, given this group's track record with their earlier poisoning effort (Glaze) vs. the adulatory press that they managed to arrange, I don't expect much from Nightshade despite the similarly adulatory press. Great media relations game, though.
https://www.nytimes.com/2023/02/13/technology/ai-art-generat...
"What if I take my own pictures, and train the data on my own pictures?"
Their answer:
"No"
Luddites, nothing more. They will become obscure while everyone else advances. She uses her colored pencils, I use digital art. Guess who works at a coffee shop.
> Guess who works at a coffee shop
Her? Happy and fulfilled as a human being no doubt.
1. Run a large number of verified unpoisoned images through the AI poisoning algorithm
2. Create a large high-quality data set consisting of the before and after poisoning image pairs
3. Train an AI using this data set so it can detect or even reverse the poisoning
4. AI train stops for no one
But that's about it, and this isn't meaningfully going to impact AI image generation at all.
But now people are posting images instead of text. And it all looks so stupid that it only deals reputational blow to some subgroup of artists who now look like a group of technically illiterate luddites.