Glaze: Protecting artists from style mimicry
glaze.cs.uchicago.edu
glaze.cs.uchicago.edu
This is a hedge against courts deciding scraping data for training purposes is valid.
Maybe you're allowed to scrape data, but with this, now you are applying a filter (creating a derivative work) to defeat a copyright protection mechanism, both clearly prohibited in current law (US jurisdiction at least). For any serious player scraping this opens your buisness up to huge lawsuits. For any serious player making tools, you'll specifically avoid defeating these techniques. For any minor player you'll now have to go to the backwaters of the internet for tools to do this that you hope won't steal your bitcoins.
Every notable artist will be only upload their art to sites that offer something like this, paid at first, but when the cost are low enough, pretty much every site that wants art content will offer it.
This isn't a technical solution to this problem, it's a political solution that happens to use tech.
It's a political solution that capitalized on fear, yet does not offer tangible protection. Even watermarking is more effective.
The point for a given artist is that your images aren't there so the AI can't imitate your style (at least not by using your name).
It'll work regardless. All it needs is X artist uses this medium, does these sorts of pictures, looks like artists Y & Z, etc. etc.
SD does okay on some artists that don't have works in the image dataset. All their name does is pick out a position in the embedding space.
I don't think this is true. Can you provide an example?
But i do feel that unless your style is very unique and no existing "roots" in existing styles, it would not be possible to "protect" it technologically.
That's the purpose of the training, no? To generalize the model so that it can produce images it has never seen before. That includes images in styles it has never seen before.
Whether or not models have generalized to that point is a different question, but if they do, (and let's be honest, alot of artistic styles aren't that unique or different) the only thing that would be different is that the model cannot conjure up the style by providing the artists name in the input, instead one would have to describe the style in other ways.
So CLIP can have picked up an association that a named artist usually is synonymous with for example "broad strokes, moody lighting" and then that is fed into the diffusion model, which doesn't know the artist but DOES know what broad strokes and moody lighting is.
But sure, if CLIP didn't know about the artist name either it won't work of course.
By the way you can still just enter the particulars of the artist you want to mimic by text as well. There is not THAT much information in a style and you won't need to feed an image into the system.
I guess all of this with artists trying to protect their online works by watermarking or glazing will only be a very short speedbump for better or worse. If a human can do a 1-shot style transfer by a single glance at a work, the next round of AIs will as well, and won't be hampered by adding noise to the works and you might have "style extraction" tools that could work like chatgpt in that you iteratively instruct by text commands what to do to get closer without ever letting the AI look at an image.
> There is not THAT much information in a style and you won't need to feed an image into the system.
I think it really depends on the style. Some styles are simple, others are more complex. How that translates to an AI understanding it though, I have no idea. But in terms of brush strokes I don't think it's fair to say every style can be described as "broad strokes, moody, dark". Some very simple styles, maybe.
Aren't you kind of jumping ahead and assuming copyright protection applies to ML training? I thought this remains undecided right now.
In short: "Section 103 (17 U.S.C Sec. 1201(a)(1)) of the DMCA states: No person shall circumvent a technological measure that effectively controls access to a work protected under this title."
So if Glaze is viewed as a protection measure, circumventing it would itself be a breach.
Hmm. That's a good point and now i am less sure.
I suspect it would come down to intentionality.
if f(x)1011 and 1010 both "look" the same shouldn't a well designed AI learn equally well to recreate that particular look. It doesn't need a "deglaze" circumvention step it just needs a continually improving learning process that automatically accommodates both glazed and un-glazed images.
It does, though.
> This is a hedge against courts deciding scraping data for training purposes is valid.
For that hedge to work, though, it needs:
(1) Not to substantially degrade the art it is used on, and
(2) [To protect anyone other than a major player against other major players] ] Not to be trivially bypassed by in a way that can be incorporated in automated workflows.
> now you are applying a filter (creating a derivative work) to defeat a copyright protection mechanism, both clearly prohibited in current law (US jurisdiction at least).
If the Fair Use exception applies to training an LLM (which is creating a derivative work, itself, before considering Fair Use), then its extremely clear that applying a filter to incoming works as part of that process (even if not if the filtered work was used for any other purpose) will also be protected. So the derivative work thing is useless.
The circumvention measure thing might technically work (in that it interjects a violation into a workflow that would otherwise be Fair Use), but as a practical matter that doesn’t matter for most users or against most violators. Moreover, to the extent its primary effect would be to adversely impact otherwise noninfringing use, it would be a textbook case of a justification for the Librarian of Congress issuing a DMCA exemption, which would then negate the legal utility entirely.
> For any serious player making tools, you’ll specifically avoid defeating these techniques. For any minor player you’ll now have to go to the backwaters of the internet for tools to do this that you hope won’t steal your bitcoins.
The proof of concept defeats already demonstrated use…the same tools that are used for AI image generation and training models on artist styles in the first place.
> Every notable artist will be only upload their art to sites that offer something like this
The samples I’ve seen of the damage this does to art suggests that this isn’t the case.
> This isn’t a technical solution to this problem, it’s a political solution that happens to use tech.
It’s neither, its just a non-solution.
That's a big if. Sure the librarian of congress could justifiably issue an exemption, but i think it is far from garunteed that they would.
I think you are underestimating what a cooling affect FUD related to anticircumvention could be. Just consider all the stuff related to dvds that went down back in the day e.g. Dmitry Sklyarov being arrested at defcon (even if it didn't stick). The uncertainty could definitely have a major cooling effect.
I think you are overestimating it. Because you overestimating the degree to which the present legal context is similar to when DeCSS was an issue, and because you are underestimating the degree to which the financial and political power imbalance between the xxAAs and the other side was key to the effect then, and because you are misunderstanding who is doing the thing this most effects, before countermeasures (its not the easily deterrable parties), and because you are (based on the circumvention POCs) overestimating the degree to which any controllable specialized circumvention tool is needed to circumvent this type of “protection”.
I agree that power imbalance is a significant factor that is different. Fair point.
Why would these parties not easily be deterred by legal concerns? Well there are a variety of groups in this space, many of them are corporate/start ups, and usually they are more adverse to legal risk than your average pseudo-anonoymous citizen.
The statute is essentially "we don't care if what you're doing is legal or not and we're not going to wait for a court to decide, we're playing this trump card to make what you're doing illegal regardless of it's copyright status."
It's equally scummy whether it's over AI training, preventing someone from using material from a DVD under fair use, or suing for Joe Blow for refilling his inkjet cartridges.
No, in none of them are the victims mainly large corporations.
In two out of the three, the scumbags who are victimizing are acting through a large corporation, but I’m not sure how that is material.
Do you not think that the response is justified in the face of the livelihood of artists being actually stolen from them as we speak?
Copyright maximalism makes the creative world worse for everyone in the long run.
Artists don't need to collaborate with the tool running their livelihoods either.
Ruining the lives of actual creatives is making the creative world a worse place right now, so spare me the thought for the long run.
Besides a diffuser trained just in public domain is going to be as dangerous to their job, they seem to generalize well enough, and if their style is anywhere close to anything existing pre 1920 they will have ai spitting out the exact same thing they do and no recourse.
And to get to the point, actual creations can use these tools to increase their output then fold or more. A reduction in commission price is to be expected, as is expected for them to use the tools to produce more.
They are not outcompeted by ai, they are outcompeted by a new crop of artists bracing ai tools to work at a greater volume and thus can sustain a lower price point. More art will be produced, not less, and those in true danger are the artisans that aren't embracing the ai industrial revolution.
Seems tech won’t be satisfied till its devalued the entire human experience of growth and effort and replaced it with quick answers and cheap results.
Edit: Then one day when the species goes extinct because we’ve got our robo-waifus we can exclaim in our final breathes that the luddites were wrong. All technology is great, no issues whatsoever.
Edit2: I’m kidding of course. We’ll either automate baby making and/or have the means to live forever so there’s no incentive to reproduce. Silly luddites and their antiquated notions of human relationships and death. What will they complain about next, human cloning or brainwashing? They’ll probably bring out some trope like “stop playing god” or “stop bending people to your will”. The technology exists so it must be used. If I don’t brainwash people someone else will right? The ones who fail to adapt to the changing world will just have to do whatever I say.
A rising tide really does lift all boats, whether you believe it or not. History furnishes no exceptions to this rule.
To those looking at becoming politically active in this area, it will be much better to work to democratize AI than it will be to try to stop it. The former will be difficult, the latter impossible.
It doesn’t, though. The models that are transforming the markets now would continue to do so, in much the same way, if no additional art from working “traditional” artists was used to train either base models or fine tunes.
Sure, if this existed before the models – which it couldn’t, since it incorporates one of them directly – and had been applied to all existing (including public domain) art and it was completely effective and had no immediate countermeasure, it could have obstructed or delayed things, but now its trying to unring the bell, and it can’t, even if it worked flawlessly on a technical level.
> The robot can't train on its own output because it doesn't understand what it's supposed to be drawing.
Actually, a very significant source of training data for new training of models is…output of existing models curated by the trainer. The robot can be trained on its own output.
> Artists don’t need to collaborate with the tool running their livelihoods either.
Ironic statement, given that they do even to use Glaze, which incorporates exactly the tool they are blaming to do its work. In fact, the entire poisoning attack is done by injecting other artist’s styles into their work, in the hope of confusing style extraction. That is, the tool literally does exactly the thing it is intended to prevent, as its preventive measure.
To be fair, the economy is in a dump. Reduced commissions and positions being cut might have more to do with that than AI. That said, the market for “generic” art is probably going to be taken over by AI.
Nothing is taken over by AI, but just young clinical psychopath types are being enabled to be burned later at the expense of all. That's kind of wrong.
Also - if I want to get, say, cover art for my book - why should I hire a human if Stable Diffusion works just as good? I'm also not interested in ripping off a particular artist's style. Do those artists really think that I'm going to hire them after they opt out of being included into ML datasets? Nope - I'll just experiment with SD a bit more to get the style that I want.
Are you sure it works like this? Because SD training usually starts with resizing, cropping, and encoding to latent space. If applying a filter is clearly prohibited, the current approach of "resizing, cropping, encoding" is surely too, right?
b) The DMCA is a US legislation, what about training done in other countries?
Maybe? It certainly prevents reusing the work in that specific manner without authorisation. It doesn't prevent just seeing the work though, so it's a bit up to interpretation.
To b), US-like copyright law effectively applies in any WTO-conforming country. Anything that implements WIPO functiouns roughly the same way. Fun DMCA fact, there is no fair use provision. Any use is likely to be criminal. Another fun DMCA fact, apparently ripping out spyware and republishing is entirely legal. Huh.
> It certainly prevents reusing the work in that specific manner without authorisation.
Well, technically speaking, it doesn't prevent it. It just messes up the results of the work being used in that manner. And what if someone builds a training workflow that can just ingest such changed images and use them without being negatively affected by the changes?
> so it's a bit up to interpretation.
Interpretation that would likely have to be decided in court, and likely in a very drawn out and very very very expensive manner, with uncertain outcome.
A machine-readable tagging that simply says "noone is allowed to use this for training AI" sounds way easier to argue in court to me.
That is what encryption does as well. You can certainly attempt to watch DVDs without permission. It won't be very enjoyable. And what if someone builds a viewing application which can just watch those DVDs anyway? You see, if this is legally protected, building that workflow is circumvention and very, very illegal. Defining "effectively restricts" is left intentionally up to interpretation because there is no clear line between messing up the result and preventing access.
Encryption prevents usage of the data. This doesn't. The data can still be viewed without any special software, device, password, key, etc. A sufficiently robust ingestion engine could still use it for training. In fact, even an unprepared engine can train on it, it only messes up the outcome.
Honest question: Is it harder to argue legally that I DRM-protected my work if I publish them in a form that needs an encryption key/software/device, or if I publish them for everyone to see after changing some pixels around?
> if this is legally protected
"If" is the important term here. It was mentioned above that "maybe" this counts in the same way as a Copyright protection measure.
I don't argue against that. Maybe it does. That is for lawmakers, courts and similar legal experts to decide.
My opinion as someone who isn't a lawyer, is that it would be EASIER to get courts to agree on that machine-readable tagging, simply disallowing usage of works for training, is similar to DRM measures, and ignoring them should be punished the same way as circumventing copyright mechanisms.
The added bonus for users: Such tagging is easier to implement, easier to update, there exists prior law already covering it (see several european countries) providing legal guidelines. Plus, artists wouldn't have to mangle their works to implement them, and it is useable with all forms of data, not just images.
It should be. It'd be really great if it was. Sadly, lawyers wrote the DMCA. It has to be a measure which actually restricts access in an effective enough way, just saying "don't touch this" isn't a measure because it doesn't effectively prevent the usage. 17 USC § 1201(a)(3) and all that.
If training on image sets isn't copyright infringement, "don't use this" doesn't count. It's a license, and you're not infringing on it. If it is copyright infringement, "don't use this" is the default and you can't use anything without explicit permission, effectively requiring datasets to only include CC0 images. Since the first one is way more likely to be true, you instead use the 17 USC 1201(a) which prevents circumvention of technical protection measures, by creating what is hopefully a technical protection measure. Is it foolproof? It was never meant to be. It's an attempt at best. But it's better than relying on a law which is incredibly likely to never apply to dataset scraping.
Preventing the usage of the data for a specific application vs total restriction is the big issue here. Is it enough to qualify as a technical protection measure? Maybe, maybe not. The courts may agree, or they may not. But it's fairly established in conversation that dataset scraping isn't infringement, so tagging it doesn't really work.
Conversations aside, laws can adapt to include terms regarding tagging. As said before, there are legal examples for this, eg. in the EU:
https://discoverdigitallaw.com/is-web-scraping-legal-short-g...
Quote:
the new law, that must be applied by all EU countries until 7 June 2021 (Directive (EU) 2019/790 on copyright and related rights in the Digital Single Market or ‘DSM Directive’), in its Article 4 provides an exception from the rights of the database owner mentioned above in case of ‘reproductions and extractions of lawfully accessible works and other subject matter for the purposes of text and data mining’ unless ‘the use of works and other subject matter referred to in that paragraph has not been expressly reserved by their rightholders in an appropriate manner, such as machine-readable means in the case of content made publicly available online’.
Again, I'm not a lawer, but to me that seems like it's up to lawmakers to do their homework, and update existing laws to deal with the reality that
a) data mining exists and is useful for lots of things
b) people want to make their works available publicly, and therefore ...
c) people publishing works need a workable, stable and reliable way to tell others whether they are okay with their work being scraped and used for analysis/training/etc. or not
And as I said above, ideally such a solution doesn't require changing the published data in some way, and works for all kinds of data.
The cloak is absolutely meant to prevent copying of the image. Thd cloak is only effective against ML, not humans viewers.
When CSS encrypts a DVD, it scramble the bits so your unauthorized computer can't decode it. You could decrypt it by hand.
This is not a problem from the start as you are reading it but not coping it.
And as far as I know an artistic style is not just protected because someone uses a style.
There are million artists out there. Style references a certain amount of rules.
You can't just protect styles as this would also hurt artists
What surprises me most is that the paper does not consider the most obvious case: resizing of images before training; people usually train an SD model on resolutions of 512/768, so the noise is destroyed to a large extent even without realizing it. Why resizing the images is so effective is shown in this article about adversarial attacks: https://towardsdatascience.com/know-your-enemy-7f7c5038bdf3, the model after being trained with adversarial training learns independently to rescale the images as an adversarial defence.
They also not consider in the Countermeasures section the fact that people can use diffusion models for what they are trained to do, denoising the images, if you put adversarial noise on it you can use img2img to remove it (maybe even canny controlnet just to guide it even more).
Final detail (maybe) is that the paper do not address the fact this only works on models trained with a VAE and not on diffusion-only models like Dalle 2, Karlo, Imagen (maybe MJ, who knows).
The software that applies the "protection" do not run on the GPU even though it runs SD and gradient descent, so it could even take 40 mins to apply it on a single image; it also violated the GPL license of DiffusionBee: https://www.reddit.com/r/StableDiffusion/comments/11sqkh9/gl...
In case it's necessary to remove the adversarial noise, the simplest way: https://github.com/lllyasviel/AdverseCleaner (16 lines of Python)
Are there plans for any DiffusionBee contributors to sue or file a DMCA takedown against Glaze?
Edit: https://mobile.twitter.com/ravenben/status/16364393355693752..., he is talking about releasing the code behind the front-end even if the back-end also violates the license.
GPL says: if you distribute something that is based upon GPL code, then you must be prepared to offer the full code to the thing you distributed, on demand. You can even charge a reasonable fee to cover your costs.
Someone asked them for the code, they responded by publishing it, so there has been no GPL violation. There is only a violation if they refuse a valid request.
Isn't "critic" / "research" / "education" good fair-use cases?
I don't know how small the snippet has to be to be considered fair in court, but it looks like they copied large portion of code.
Fair use is violated by the scale of copy-paste involved, and well, none of "critic", "research", or "education" are the uses involved in the release of the derived art. (Writing a paper would fall under "research", iirc)
These are grad students and artists. It was probably a mistake - or they found it was covered under fair use. But, nope! Stallman wouldn’t approve! Canceled!
On closer inspection, it turns out the tool itself was plagiarized.
That was certainly a mistake.
Whether a tool prevents plagiarism has nothing to do with whether it is itself plagiarized.
Plagiarism is a misconduct whereby someone else's work is presented as one's own.
Falling afoul of the GNU Public License in some way isn't ipso facto plagiarism.
"Plagiarism is the fraudulent representation of another person's language, thoughts, ideas, or expressions as one's own original work." [1]
Tool claims to prevent plagiarism, paper uses "plagiarism" language throughout. [2]
Writers of tool used substantial portions of a different tool without acknowledgement. Source explicitly calls this out as plagiarism in first paragraph [3].
[1] https://en.wikipedia.org/wiki/Plagiarism
[2] https://arxiv.org/pdf/2302.04222.pdf
[3] https://www.reddit.com/r/StableDiffusion/comments/11sqkh9/gl...
(edit) To be sure: somewhat later, authors do seem to be correcting their behavior, but only after having been called out. ( https://mobile.twitter.com/ravenben/status/16364393355693752... )
If you have the license to use some source code (e.g. it's public domain, or you purchased a license which allows royalty-free redistribution) then you have no copyright issue. If you submit that code to your school as your assignment, or part of your assignment without attribution, then that is plagiarism nonetheless. It's a problem between you and the school (and the world at large), not between you and the author of the code, who allowed your use.
The GNU license is not concerned with academic misconduct. Users are encouraged to use the code as the basis for making something that works for them. It has to be redistributed according to certain rules so that the freedoms originally granted to the users are preserved. The concern isn't that the correct author has been identified to the users, but that they can get the source code and build that same thing, and also redistribute it with their modifications, if they wish. If you tell the users that this is based on GPLed code from such and such, but don't give them the code, then you are innocent of plagiarism, but still in violation of the license.
Most programs use code that their author didn't write---often, a lot of code; that is understood. When I see an announcement like "Glaze: Protecting artists from style mimicry" and read its synopsis, I'm assuming that 90% of it is cobbed from libraries and adapations of other people's code. We don't think of it as plagiarism unless the central, distinguishing idea of the work is something that was consciously taken from someone else and presented by the purported authors as their own.
And even then, if that plagiarized idea is effective in implementing something which prevents plagiarism, that just amounts to irony. Irony is a real philosophical construct, but not a software defect.
If you plagiarize, you will also tend to automatically violate both copyright law and most licenses, including BSD, MIT, and GPL. The GPLv3 is a bit more complicated than the other two, but section 4 and 5 state that you must include "appropriate copyright notices", which means full attribution.
Here's a more in depth discussion specific to the GPLv3:
https://opensource.stackexchange.com/questions/4577/does-gpl...
[1] (edit) I was thinking that CC-0 (being close to PD) would not require attribution. But on looking it up, it's a bit more complicated.
(edit 2) In general I don't think "that is understood" holds. If you plagiarize copyrighted content, you might be breaking one or more laws. Depending on what you did exactly, how much you did, and how nasty the other party is feeling: you may be forced to take "your" work down, delete it outright, hand it over to the other party, pay damages, pay a large fine, or even go to jail.
(edit 3) I agree that plagiarism by itself need not be illegal in some contexts. But if you then distribute your plagiarized work (which is very easy to end up doing these days), you do end up infringing copyright law.
The USA copyright law doesn't mention the word "plagiarism", though there is a small section in it entitled "Rights of certain authors to attribution and integrity". I'm pretty sure FOSS licenses are not relying on this. They rely on the idea that the license lapses if any of its terms are not met. When the license lapses, then the right to redistribute is revoked. A lawsuit revolving around someone removing copyright notices would most liklely be based on unauthorized reproduction (the license to redistribute is not granted to someone who removes the notices), not violating attribution rights (license to redistribute is unconditionally granted, but authors retain attribution rights, which are being stepped on).
Software licenses as a broader class not limited to FOSS do not necessarily discourage plagiarism. In proprietary software, it's common to be able to license a software component on such terms that its authors need not be mentioned anywhere; neither the binary build of the program, nor any accompanying documentation have to display any notices related to the licensed part. If you license parts of a program in this way, and say that you wrote all of it, that is plagiarism. In an academic setting, it could get you disciplined and kicked out of school.
Homework writing services probably license their solutions on similar terms (out of necessity): you may present the purchased solution as your own work.
Free software writers are not mainly concerned with plagiarism. You're encouraged to take a program and make it yours, if it doesn't work the way you want it. You may add your own name to the copyright notice if you make more than just trivial changes, and if you add new files that are not a derived work of the program, you can use your own license for them, provided it's compatible.
Someone who, in redistributing some code, neglected some requirement stated in its free license isn't ipso facto a plagiarist: i.e. isn't guilty of fraudulently presenting the entire program as their sole creation. Situations in which that is the case will be painfully obvious.
Therefore, in summary, it's basically just trolling to throw around the word "plagiarism" when someone made a mistake or oversight in some redistributing use of a program.
What possible good itentions could their possibly be here?
Were they unaware of the idea that you shouldn't wholesale copy other people's work? That seems impossible by their own admission.
Were they unaware that they were using the software? That seemes implausible.
Did they believe that rules only apply to other people and not themselves? If so that is my definition of a bad person.
Quite frankly i don't see any possible way this could be dismissed as an innocent mistake. Arguably the hypocrisy makes it much worse than your average gpl violation.
Artists and grad students are people too. Being an artist doesn't alleviate moral culpability.
Edit: however to clarify, since you said people wanted to "ruin their livlihoods". I want to be clear that i don't support anything like that. Negative consequences should fit the "crime", and ruined livlihood would be way out of porportion.
...and as soon as this was brought to their attention, the authors took out the infringing code and reimplemented it from scratch.
However, I'm surprised at how little empathy is being displayed towards artists by so many here. Having a model train itself on your work is not the same as having another human being inspired by or even copying your work, despite the fact that both involve some sort of learning.
I am not an artist but I build all my solo programming projects "in the open". My programs are mostly niche stats models, and frequently people message me with questions and ask for help integrating my work into their software. I'm happy to do so. I don't care if they credit me in their final product, but knowing that humans went through my work and took time to appreciate it is a big motivator for me. I would imagine that many artists feel similarly.
On the other hand, knowing that OpenAI, Github/Copilot, etc, train models on my work and turn it into some pay-to-play API, without a human ever seeing my work during the whole process is a nasty feeling. At that point, I've just been turned into faceless cog to generate profits for big tech shareholders. Luckily, these are just side projects for me and I can just make these repos private, but of course artists are forced to "build in the open" by the very nature of their work.
I understand the disappointment of real artists that the tech bros are stealing their lunch, but sadly it's the way all these things go. Pandora's box is opened, things aren't going to go back the way they were.
Even if an artist manages to completely guard themselves against exploitation by AI, the market expectation for art commissions and work will be 100x higher than it was before AI art in a few years. It's like the scene in There Will be Blood, and the milkshake of traditional artists has been drunk already.
1.Like climate change, its a commons issue. AI learns from the artists at large, so a single artist protecting themselves does nothing. Nobody copies specific artists except the very top tier who have distinct and beautiful styles.
2.AIs will still have all the art produced before 2023 to train on. Which is a gigantic amount with huge room for optimization in training.
3.This can be trivially circumvented. Given the requirement of the glazed image to look identical to the human eye, it must be possible informationally to reproduce the image without the glaze.
4. It should be trivial to train a quick GAN to revert this glaze, given you can trivially artificially create before-after datasets using this very tool.
I think the uncomfortable answer is that for a lot of commercial art uses, nothing changes. In many cases people just want something to fulfill a need and aren't really all that picky about it. Maybe the style changes to be one of the Old Masters instead of something current. Getting it fast and cheap means they don't have to think about it much. A great many commercial art jobs will vanish just as they will outside this hypothetical.
Perhaps we should pause and identify what outcome we want before prescribing policy. Are rights our priority, or are we trying to secure the incomes of artists?
So I get why it feels wrong to many, but I also have no trouble placing it and those objections in a context of a recurring historical pattern.
Compassion is not a policy. Compassion is not a plan. Compassion isn't what coal miners or Luddites wanted. What they wanted was to freeze in amber a way of life that served them well and could be passed on. That's both a reasonable thing to want and a very unreasonable thing to expect.
Or maybe you don't know what you want as an outcome. That's OK. It might be worth thinking about that.
For my own part, I'm not too worried about programmers quite yet. After all, programming is the easy part. All you have to do is get the business to decide precisely what they want and communicate it clearly. For the future, well, adaptability and rapid learning have been the hallmarks of every good engineer I've ever worked with. We're a flexible lot.
For sake of clarity I am an engineer as well. I hope you are right? As I see it now, it seems like every single job is on the chopping block before I hit retirement age. We already see a few experts arguing about this already.
I think I am doing a poor job conveying what I want or mean by compassion, as you said that is true that the Luddites or coal miners in those examples want to pause time. That is clearly a very bad idea when it comes to something like coal, but at the same time I think when I say compassion it is not that I want to tug at the heartstrings of ourselves and others. I mean it in a selfish way that when we are able to answer the question as to what people do when their livelihoods are removed from them, we get a better functioning society. When we use coal miners as an example there has been a few studies that have shown much higher usage of drugs as despair has grown in those communities. I see that and have extrapolated that out to the large white-collar hypothetical hordes of unemployed commercial artists. Perhaps all desk workers (my self included).
It legitimately has kept me up at night trying to think what sort of policy, what responsibility do we all have now? Economically nobody owes anyone anything, but is that an ethical answer?
I don't know what that looks like. Which goes back to my original answer, I don't know what I want as an outcome. I just think that perhaps, we will see less animosity and less anxiety when we have better answers to our modern Luddites.
Therefore, why is it not the artist's responsibility now, to look for alternatives, and not wait till they truly become obsolete, and personal resources run out? If you went back in time and told the luddites that in 5 years time, their services would no longer be required by society, would they not retrain themselves instantly, instead of waiting for the 5 years and hope that society has some sort of welfare program ready for them?
Their efficacy is the more interesting question to me.
The full-resolution examples I’ve seen have shown this as much more destructive to the underlying art than usual watermarks, to the point where I’m not even convinced people would use it on samples when the underlying art wasn’t displayed online but sold through other channels.
> YouTube already transcodes your videos, why not DeviantArt encode your art except to verified human accounts or something?
So the idea is to be “protected” against automated scrapers building datasets for training, say, the core StabilityDiffusion models, but not against the huge number of individuals training (and remixing) checkpoints/loras/TIs/Aesthetic Gradients and exchanging them that are driving the ecosystem?
For someone with more effort and time to dedicate to cloning a specific artist, maybe not as useful. But I know less of this type of use case/the tools in that part of the ecosystem
Look at something like Civitai [0] and the number of users submitting things there: pretty much everything there (there’s a few things like poses and tools that aren’t) is one kind or another of trained model (though some are merges of existing models and extractions of one kind of model from another.)
New Cloudflare offering:
ECA: Edge Cloaked Artwork.
Enabled at the edge with our blazing fast CDN, ECA is a brand new Cloudflare service that offers a groundbreaking solution for digital artists and creators! With our CDN network, we can now cloak artwork to protect it from unauthorized use, while still maintaining its visual appeal to the human eye. Our technology modifies the data of your artwork in a way that's invisible to the naked eye, making it nearly impossible for AI companies to use your artwork in their training samples. This not only ensures that your artwork is protected, but also helps prevent the proliferation of unethical AI practices. Trust us to safeguard your artwork and its value like never before. Try Cloudflare's artwork cloaking service today and experience the peace of mind that comes with true digital security.
- CHATGPT 3.5
Feel free to re-use Cloudflare.
Even one pass of SD's img2img with low denoise is enough to bypass this data poisoning attack. It's a useless attack that makes training a tidbit more inconvenient.
The project also gained infamy when it stole code from an open source project with GPL license without giving credits.
Additionally, SD now offers opt outs for artists and MD likely does not train on these artists at all, and there are way more effective ways to protect art than using this algorithm.
It all seems good intentioned but also naïve, like when robots take over doing all surgery and massively reduce the risk of accidental death, surgeons will get together and insist they ban the robots and that humans still do it better, even if they don't.
“The Problem with UChicago’s Glaze”
https://jackson.sh/posts/2023-03-glaze/
Yesterday, the SAND Lab at UChicago made Glaze available to download. It’s a tool to help artists protect against their work being used to train AI models. It got a bit of buzz last month, including a New York Times spot. However, it has some issues:
1. The authors plagiarized code from DiffusionBee, an AI art tool licensed under GPL.
2. The paper contains inflammatory and libelous language with no legal backing.
3. It doesn’t work and I was able to execute a proof-of-concept bypass in minutes!
Each of the sections below will go into further detail on these points.
I haven't really been following this closely, but according to that Twitter thread, supposedly that was fixed yesterday.
> The paper contains inflammatory and libelous language with no legal backing.
I don't think I can roll my eyes harder. Maybe go for RICO next, too.
I don't think that copying an artistic style is stealing legally or morally (I think the moral issue is "passing off"). But I think the reaction ("libelous language") is way over the top.
I think property rights are a sham, if we got rid of copyright altogether for code and art and everything else I'd be very happy. But if you're going to falsely accuse people of stealing when what you're actually accusing them of is copyright infringement, probably don't have your literal only production be a product of copyright infringement.
Copyright is a subset of property rights.
> Intellectual property is not property.
Intellectual property is intangible personal property is personal property is property.
> Infringing copyright is not theft.
Right, it doesn’t permanently deprive the property holder; its more like trespass than theft.
Generative art models will always have the advantage in this fight. For them, they just need to crack the protection and update their model. For artists who want to protect their work, they’ll have to continually update their whole portfolio of images and republish them whenever a new version of Glaze comes out (and attempt to ensure that their images with old versions of Glaze don’t end up in a training set, which will be difficult if not impossible).
Kind of a David and Goliath situation (yes, David is an acknowledged sniper).
How do you figure?
They’re not trying to produce sellable images; free adversarial images are useful to them.
Only if they were correctly tagged/ labelled. Of course, academics always have access to cheap labour, aka undergrads.
More seriously, reglazing old images isn't very helpful, since they will amost always be already in training datasets by that point.
1) That's not how it works.
LAION is a set of urls, not images.
Actually distributing the images would be blatant copyright infringement and, even if you don't care, is prohibitively massive.
So, yes, modifying the source images will be effective against new players who want to scrape the dataset... and yes, as the tech to 'build your own stable diffusion' becomes more commoditized, you can be 100% assured that multiple groups will be doing it in the future.
...this also has a FAQ that states even a limited number of poisoned images (eg. most recent art by Greg) can cause artifacting in outputs.
2) ...but more importantly, are you joking?
Artists complain and artstation does what, update their TOS?
That obviously does nothing.
So, companies start inventing this anti-ai watermarking tech that shops like deviantart and artstation can pickup and put into their pipelines and make into a pro-feature.
If you think it's not going to happen, you are 100% deluding yourself. The playbook is so obvious, I'd be amazed if it's not already a WIP for these places.
Artist friendly. Make money. Big players can work around it and small players are screwed out of any good images to use for training.
It's going to happen.
3) '40 minutes to return a single image...'
Oh come on. Are you objecting because you think the process has no technical merits, or because you don't like it?
Technically, this isn't a limitation. These places already have massive image processing pipelines for resizing, thumbnailing, etc.
That's because AI art was a mere research toy back then.
Now it has massive economic potential, do you really think downloading say 100TB of images is a big deal to any company? Less than $100000 to get an invaluable database that's safe and sound. Its also not prohibitive at all to just clone this data into a NAS array, and just literally ship the hard drives to whomever needing it.
And trained into models.
It also turns out to be useful in often helping human beings coexist relatively peacefully and sometimes negotiating mutual interests productively, so like a number of things that don't exist until enough people agree to conduct themselves as if they do, it's probably a good idea in general.
Intellectual property has some distinct characteristics, but it's fundamentally the same.
And assuming entitlement to any available data for ingestion into a training set without a moment's regard to the labor and agency of the people who produced it has things in common with the ethic of a thief who, coming upon your stuff, decides that if they can take it, it's theirs, society's construction of property be damned.
(... I'm literally wearing a "YOU WOULDN'T REIMPLEMENT AN API" t-shirt as I type this.)
The similarities matter more than the differences here.
And the problem really isn't breaking new ground in general. People saw the similarities at least as far back as the printing press. The problem is getting people to care about the claims others have on the fruits of their labor, and why incentives might matter for everyone.
The invention of the printing press didn't stop writing, the invention of the camera didn't stop painting, and the invention of Stable Diffusion won't stop creation of novel art. But it definitely upends a known methodology for extracting value from non-fungibility of labor product.
In any case, the labor-to-value question is moot because Stable Diffusion will enable companies to generate unlimited visuals off the single-time-compensated manual labor of a handful of artists (for no other reason than some artists will take that deal). So whether any one artist is in the training set will become irrelevant as people choose nearly-free good-enough product over far-more-expensive handcrafted product most of the time.
Glaze is useless when an advertising firm is fairly-compensating consenting artists to toss their work into the meat grinder and that grinder then still churns out work that swamps the individuality of traditional-methods artists.
Well then, if the fruits are worthless, then presumably there's no loss from excluding it from a training set, and creating norms or even laws which allow people the privilege of negotiating the basis on which their work can be so used, right?
Of course, nobody believes that. Much like no one really believed that unauthorized copies of 18th century print works were valueless just because the production could be industrialized. The enterprises producing unauthorized copies did so because they knew full well they could capture the value... without any of the pesky pro-social obligation to respect the labor and interests of those who created it.
The interest in the works as training data betrays the position. This work is valuable, and works derived from it are valuable.
> Glaze is useless when an advertising firm is fairly-compensating consenting artists
I imagine there will be artists who opt in after negotiating the basis on which their work can be used -- although who knows what kind of surprises there could be there, some might even demand equity or collectively bargain.
But deciding there's no need for a framework of negotiation and consent just because some people will get to yes sounds like a really bad precedent.
And if the argument is that this isn't sufficient to protect interests, well, sure. Necessary but insufficient points exist all the time, usually the thing to do is pair them with other axioms / initiatives.
The printing press didn't end copyright, but it obviated the need for hand-illuminated manuscripts. I expect automatic image generation will do something similar to hand-created from-scratch art. And yes, if there's a framework to be created to allow artists to indicate that they don't want to be involved and wish to lead the charge into irrelevance, it should be created. It's best to have a revolution be tidy not messy.
The decades of GPL violations have taught us the only way to get commercial interests to not abuse copyleft licenses is to force them via the courts, and in doing so dis-incentivize other businesses from violating the licenses.
But it isn't 100% effective by any means, and usually relies on a large commercially successful organization to already have aligned interests in enforcing their GPL/copyleft licenses.
Anything else is just ethics paper fodder.
The tech industry is full of examples.
>Once you add a cloak to an image, the same cloak can prevent different AI models (e.g., Midjourney, Stable Diffusion, etc.) from stealing the style of the cloaked image
But my understanding of what they're actually doing is a data poisoning attack on the fine-tuning process (e.g. someone tries to finetune StableDiffusion on a handful of a particular artist's images). Does Midjourney offer that sort of finetuning at all? And isn't it misleading to say that you're preventing Midjourney and Stable Diffusion from "stealing the style"? That seems to imply you're also poisoning the regular training process.
I'm also very unconvinced that this offers any meaningful protection in practice. The lesson from years of research on adversarial perturbations has been that it's very easy to make your method look successful in your own paper against a naive adversary, and way harder to make something that stands up to an intelligent counter-attack.
I'm not really convinced it's ethical to present this in the way they do to artists who probably don't understand the technical details. Even with their disclaimers and limitations section, this website gives a much rosier picture of the tool's effectiveness than I think it justified. If an artist is concerned enough about this sort thing to want to use this tool, I think they'd be upset to learn about it's true (in)effectiveness, and once they've chosen to post their Glazed images, it's too late.
The project also gained infamy when it stole code from an open source project with GPL license without giving credits.
Additionally, SD now offers opt outs for artists and MD likely does not train on these artists at all.
It is all you need in practice otherwise watermarks wouldn't be a thing.
You don't need to run faster than a lion, all you need is to run faster than the gazelle next to you.
No, its poisoning core model training as well, because the core models are trained with scraped art with artist and other metadata, which is why you can use “in the style of…” prompts for lots of artists with the core models. It also applies to fine tuning, etc.
At the very least, I'd want to see some actual experiments on training from scratch before telling artists that Glaze will protect them in that scenario. And I'm very skeptical that it would.
It's a known technique to do this kind of thing intentionally to train models that are more resistant to adverserial attacks. One reason people don't do this is the cost of running PGD is so high, but in this case your adversaries are doing it for you free of charge.
If you're an artist, of course you can make changes to your process or your content that modify how it's used. To argue otherwise is analogous to arguing we must always be mindful to stay in the surveillance cameras' view when walking about the streets.
Yes, it's a never-ending back-and-forth game that the obfuscater will probably lose in the long run (though abstractly, an obfuscation technique with >50% adoption could "win" long-term) . And yes, it's important to stay apprised about how effective such tools are.
But in the short term, the existence of these tools provides a critical counter-measure to the current narrative, which is basically that everything that can be scraped will be scraped. Returning to the cameras & streets analogy, obfuscation tools are maps that tell us about routes out of view of the cameras (even though these routes may often be blocked off or inconvenient).
Whether you hate AI art or love it, I honestly believe both sides can get behind understanding obfuscation and poisoning and making tools available: those opposed will use the tools, those who want to improve generative AI can learn from the counter-measures, etc. This kind of thing can be part of a healthy deliberative process around these emerging technologies.
AI image generators don't need to see art, they just need to know what blobs of pixels "are" in relation to words. That sort of data can be extracted from just photography of the real world - it just turns out there's a lot less of that easily available and properly tagged then art collections right now.
There's more then enough public-domain examples of "style" to do the rest (and style-transfer was one of the original AI image manipulation applications).
While that has been the tried and true formula for tech companies attitudes towards their users/customers I'm not sure that will work in this case since artists are on the supply side of the equation.
The fact that people think generative AI will improve its output by learning from the obfuscating counter-measures just drives home the point that the AI is just copying the work of others without their consent.
"This paper shows that these systems (and, in fact, any poisoning strategy) cannot protect users’ privacy. Worse, we argue that these systems offer a false sense of security. There exists a class of privacy-conscious users who might have otherwise never uploaded their photos to the internet; however who now might do so, under the false belief that data poisoning will protect their privacy. These users are now less private than they were before."
Paper: https://arxiv.org/pdf/2106.14851.pdf Fawkes: https://sandlab.cs.uchicago.edu/fawkes/
What's annoying about the AI art zeitgeist is how dramatic and frankly mean everybody is about it, on both sides. And then the sheer ignorance about how AI models work, again, from people on both sides of this stupid internet fight-of-the-month (year?).
It's always something on one side of "AI art is the future and artists should stop whining that they're obsolete!" or "AI tech-bros are literally stealing in a way fundamentally different than all prior art consumption!".
Everybody is talking past each other. Using (or not using) an ML model to do anything nowadays is politicized like mask wearing. Glaze is, at its core, a GAN. The key word is "adversarial", and it will be defeated in... already, looks like.
As for the rant: the AI art horse is being beat to death, and it's boring to see the same arguments made past each other over and over. Sorry, sometimes it feels good to rant.
No matter where you land on this issue, you have to admit that there is a credible chance of significant societal changes in the near future. But instead of discussing how to best manage these changes or how to support people who may be negatively affected, we waste our time with petty squabbles.
The artists are saying "stop" and the AI side is saying "you can't make me". One side would love a pause to have a productive discussion, and the other side doesn't want a pause at all.
The other side are human beings.
If you don't immediately and intuitively grasp the moral weight of drawing comparisons between animals and humans in this context, perhaps you can at least understand why you're hated for it. "Let's just screw over everybody who we can grind between the gears of our APIs because ha ha, they can't stop us" is not a recipe for a functioning society.
> The other side are human beings.
No, both sides are human beings. They are human beings using different tools in the same domain, and the side that wants to prevent the adoption of newer tools rather than adapt to them is…well, historically, that’s not a winning position, basically, ever. (In broad societal terms; if you want to create your own isolated subculture where certain newer tools are excluded, that can work.)
I think the arguments in this thread for why this is important, by supporters of it, indicate this is wrong. Yes, directly, this aims at preventing style fine-tunes from work of the artist who uses it (even though styles have never been proprietary.) But if you look at the arguments about why this is significant, its not anything connected to the copying of individual styles, but to the way the explosion of capacity of AI art generators is transforming the market. That’s the target. Even if a magic shield protecting every work that the creator wanted to protect from this day forward against style fine tunes (or AI training more generally) wouldn’t deflect that trend even a tiny bit – which it wouldn’t, IMO – that’s what this is aimed at sociologically.
I'm not an artist and I'm excited to use AI art generation to realise visuals I can obtain affordably. I just would like to be able to do that ethically.
The problem with the current discourse is that we're hung up on things that are largely irrelevant to providing an actual solution.
Arguments like "my IP is being violated" completely fall apart under any scrutiny and are readily abandoned in favor of some other useful proxy-concerns like "It's just copy and paste, not art!". I refuse to believe that the art community is stupid enough to make arguments like that without some existential dread clouding their judgment.
Do you really think anyone would complain if AI art didn't threaten their livelihood, in a space that usually celebrates inspiration and recognizes that IP rights often only benefit the biggest players? Do you really think a significant portion of that group would advocate for the strictest possible IP laws, which will wipe out a significant amount of human art as well, without some deeper motivation?
The ultimate issue is that people are losing their livelihoods and the ability to engage in work that is meaningful to them. We need to develop social programs that ease their fall, not run circles around the emotionally satisfying arguments that would "really show the other side what's what!". Ironically, a bunch of people are reducing one side to money hungry thieves and the other to lazy luddites below a comment calling out this exact behavior.
This is what gets me about this whole argument. When you're agreeing with The Mouse about IP law, you know something has gone awry. In reality, we should remove arcane IP restrictions, not continue to add them further into society. Copyright is already, what, 100+ years on average?
> The ultimate issue is that people are losing their livelihoods and the ability to engage in work that is meaningful to them.
Indeed, people are using a moral argument to mask their deeper intent, their fear of the economic damage done to their livelihood. In this case the solution is not to further calcify their work, but to...solve the actual issue, via UBI and the like.
Are you going to buy the necessary number of congressmen to make this happen? Because the people being aggrieved certainly can't afford to--and so "well just pass UBI" is the "just draw the rest of the owl" flavor of not-helping.
Insofar as its about treating “style” as a proprietary entitlement against the rest of the world, despite that it has not ever been proprietary, a kind of entitlement to an extension of copyright which itself is not even notionally a matter of fundamental right but privilege granted in the expectation of externalized utility, I am not convinced that it is a legitimate grievance, even to the extent it is clearly a genuine one.
> Arguing that they’re only saying that because their jobs are being affected is, well, not the best looking argument I’ve ever seen.
I didn’t argue that. I argued that the motivation is not that the threat to their jobs from the thing this would correct even if it worked, but the threat from something which it does not, even in the most optimistic view. That is, yes, its because “AI art” threatens their existing mode of work, but not because “style transfer based on their current and future artwork” moves the needle, in any meaningful way, on the threat. It is fundamentally (whether they are entitled to their existing mode of work or not) misdirected for the motivating concern.
I don’t think style is even the main issue here, although it’s a significant one. You may be right. Without scraping vast troves of copyrighted art, and using it in a way that is not at all clearly fair use, these models would be nearly as good at almost anything. But if that’s a legitimate concern, the fact that it might severely hamper these models is the right outcome. If these companies want to use copyrighted works to train them, maybe they should need to license it.
> historically, that’s not a winning position, basically, ever
In what universe do you think people who are staring down the gunbarrel believe they have any chance to win at all, in anything? When “join up”, if one is even allowed (and let’s be real, they’re not being invited) is also losing?
We are talking about telos here, and tech’s incessant drive to smash it. And this industry offers no alternatives or even the barest minimum of safety to anyone.
AI art literally relies on artist's output to train. It's not horses vs the car, it's horses vs a motor that grinds up horses to generate power.
As the sibling comment notes, only one side here is saying "stop" and only one side is saying "you can't make me."
Your comment was the exact kind of argument I was writing about. You took the emotionally easy path of thinking about a large group of people as some evil cabal that has the power, unity and desire to destroy the livelihoods of others, instead of actually assessing their arguments.
People who are saying that AI can't be stopped are referring to the fact that even if they personally stopped using it - there would be millions of others who would not have such qualms. If OpenAI shut down tomorrow, new companies would open up the next day. If the government banned all AI, other countries would welcome it and start eating said government's lunch.
Striving towards better social programs is the best way forward, and UBI will be thought of as impossible until a large enough group of people campaign for it. If you disagree with that, then argue against it, but don't construct some boogeyman version of the other side in order to dismiss them.
Yes. Literally yes. It sucks. I try to minimize purchases, I research the most ethical sellers I can within the bounds of available information, and I try to reuse as much as is possible.
> That's the main issue at play here and needs to be remedied by fair and empathic argument - not by railing against a group of people who you paint as unequivocally evil.
When they stop acting like their mission in life is to drive everyone else into precarity and then to ruin, or even at least stop acting like it’s so horrible that people in their gunsights are upset about it, I’ll stop drawing from their actions the obvious conclusions.
UBI is a great way forward. Never said it wasn’t! But. It is clearly not on the radar of the people bankrolling this. If it was? They would expend the money and the political capital to do it. They would be on that goddamned stump. But Sam Altman isn’t out there buying congressmen for UBI. Could! But isn’t. So why would I believe that it’s any kind of priority?
But okay, that’s uncharitable, I’m sure. Altman is busy. So show me who should convince me that there is an argument at all beyond “you can’t stop us”, could you? Show me who is making fat stacks off of the AI craze and is also doing the political work, won’t you? Show me who is going “I am now making bank and I want us to give back structurally.” I have looked. I don’t see them. I want to push those messages. But I am reasonably confident they don’t exist. I’d like to be wrong.
I never said that there is any expectation of putting the genie back in the bottle, either. We do need political action, we do need to be emphatic and we do need to push for solutions (and UBI is a good one), and in so doing we might even fix the problems they create…but that doesn’t change their souls. It just means maybe we can work around them.
If you genuinely have a bad conscience about the things I outlined, then you are in the vast minority. 99% of people in developed nations don't care about the sacrifices made for their lifestyle, they feel no compulsion to give back or prevent the abuse of others in the name of profit. They may pay lip service, but they won't sacrifice anything to practice what they preach. Reading about modern slavery in cobalt mines will make them feel bad temporarily, but within days they'll be right back shopping for a new smartphone.
That is the unfortunate reality - our intuitive sense of injustice is strongly biased towards locality. We care when something bad happens to the people near and dear to us, but suffering that is far away or somewhat hidden doesn't really register.
We need to face this reality and work with it, instead of getting heated and self-righteous. Sam Altman's soul is neither worse nor better than the one of your average middle-class American.
Instead of assuming evil intent, assume ignorance. Make convincing arguments, highlight the suffering, offer possible solutions - these are the actions that likely lead to improvement.
And to just add a small reminder: AI isn't all bad. It's a disruptive technology, with great risks and great rewards as well.
I'll be real: the actions that are likely to lead to improvement are buying Congressmen, and I don't have the money. The AI crowd does, and they are positioning themselves to collect ever more of it. If they are in fact Actually Worried about the people downrange of their guns, why is the onus on everybody else? Why are you implicitly holding that normal people are unreasonable for not wanting to be fed into the gears and that it's not the gear-owners' responsibility to not feed people into them?
Don't misunderstand me--I can do a lot of stuff, both tech and not. I'm reasonably confident I'll be fine regardless; if push came to shove, I can enter my disappear-into-the-woods arc. I'm not going to begrudge those who can't refusing to go quietly, and frankly, you shouldn't either. I think you know as well as I do that this crowd isn't looking to talk to the people they seek to grind to dust. They don't think they matter, and if you seriously thought otherwise you'd have answered with examples.
Discussion is a luxury afforded only when all parties are going to be okay at the end of the day. You are, for lack of a better term, tone-policing the scared and the harmed on behalf of the wealthy and the harming. And you are clearly more capable of insight than that.
The reality is there's a whole group of people right now who's real problem is they didn't see this coming. AI, as always, was meant to destroy people in minimum wage jobs first. Because they're unskilled right? That's why they're minimum wage? Right?
But art, creativity. Machines can't do that! It's a classic science fiction trope, how could it be wrong? The irreducible nature of the human spirit is represented in art, it's not paid well but it'll totally be safe from automation.
Bricklayers, electricians and plumbers are going to be making decent money putting up buildings while the white-collar specialists get downsized due to automation. Any job done from a sitting position in front of a computer should be keeping in mind just how replaceable you actually are.
Because let's flip the question: why do white collar desk workers exist? Because we're specialists - we provide a service which is vital and necessary in some way, which justifies the cost and being done from a desk.
The consumers of those services are everyone else, including blue collar workers who are doing physical labor, moving material in the real world, etc.
So why should any of those people subsidize us sitting on our asses when AI can provide the same service cheaper?
I notice this in almost every topical discussion on the internet these days. People have seemingly lost the ability to have rational discussion.
The anger you notice is not context-free. People are shouty because:
- There are livelihoods at stake
- Worse, there are systems of respect at stake, many of which underwrite people (artists, coders, etc) feeling good about themselves. Rawls is a great reference point for this; feeling like you can contribute meaningfully (in a way that is meaningful for you) is one of the prereqs for social stability (said Rawls: https://academic.oup.com/book/32571/chapter-abstract/2703665...)
- We tech bros (or bras, in my case) do have a notable pattern of disrupting-for-profit. I mean, even middlebrow fare like _Glass Onion_ now casts us as villains at worst, and stooges and best, and it would be good to slow down for a moment to consider if they might be seeing something real about us. (Ask an NYC taxi medallion holder.)
- Creatives, especially, were already at a profound breaking point, one that is not easily communicated to those outside the disciplines in question; see https://www.goodreads.com/book/show/53935644-the-death-of-th... . It was created by, yes, our industry, and began in the nineties, before at least some of the readership here was born, and picked up steam ever decade thereafter.
Much like the early days of climate change reportage in oil/gas sector, the historic & intersubjectively verifiable externalities of our industry's (frankly incredible) progress feel ego-dystonic. It doesn't seem fair that something as fun as what we do could cause immiseration.
This --- to draw a tidy circle --- is because tech progress is part of our self-respect. And non-harming is part of most non-sociopathic self-respect, including our own. This is, in a word, a threat to our sense of self. A threat to our identity, and worse, a risk of moral injury.
'Moral injury' -- the harm to the self that comes from having done something wrong or harmful to someone else -- is one of the most profound roots of, e.g. PTSD in soldiers, and most humans intuitively avoid it at any cost.
One inefficient way of avoiding it is to avoid admitting that you did anything wrong, for example, by shouting back that you didn't do anything wrong.
In the long term, this works out about as well as you'd expect.
--
A quick closing word about the comments on the parent that advocate a sensible calm sit-down to work out solutions:
People are absolutely not going to have a quiet sit-down about how to fix this problem because we all know full well that very few liberal democracies remain functional enough to stickhandle everyone through such an enormous change. The USA, for example, can barely pass a budget; you think that you can get UBI through? Which would still do nothing about that Rawlsian self-respect, which, again, is just as significant to a healthy human as air, water, or food (and is therefore vital to social stability, in the same way that a steady supply of food is vital.)
Shouting is one of the few avenues left for self-expression, and thus, self-respect. No one wants to go quietly, and 'going' -- whether it be from a job, a career, a family, or even a home -- is quite clearly what is now in the cards for a great many.
And they are shouting at us because we do this shit on the regular. But never like this before. We went large with it. Did we ever.
And even if you're technically not actually doing anything wrong according to the law or your personal code ethics, other people might still get hurt. And then they'll clutch at anything to try and stop you.
(Excluding edge cases such as kink or vengance.)
My own code of ethics never allowed for that. However, I thought that generating pictures locally and using them in my stories, or showing off cool technology, or getting AI assistance with improving my skill at English, were not things that directly hurt other people.
Now it turns out that this is not the case, as the last couple of months have showed, and I find myself in a situation where my work, my hobbies, and my intrinsic existence are all at odds with the rest of humanity.
Something will have to break. I'm not sure what, yet. I'm tired of being shouted at, but it's too late in my life to completely rethink my skillset.
I personally don't work on AI, but I do feel betrayed by the industry that I felt so positive about in 2010 when I decided to join it. There are no easy answers, just a hard part for us to walk and learn from.
If I admit that this was a mistake, and it should never have existed, then that would also be accepting that everything which provides some validation for my existence is a mistake, and I shouldn't be doing that. Yes, I accept the irony, but since I'm unwilling to cease existing, I cannot pick that path.
Funnily, however?
My interactions with the art community have been two-fold. First, people who agree that it's cool, and who enjoy the opportunity to add elements to their art that they couldn't have before; whether that be writers getting illustrations, or artists seeing their works in other styles, or... a whole range of options.
Then, second, there's the people who start shouting at me and calling me evil when I try to bring gifts. Who jump into the personal insults when I'm trying to correct misconceptions. Who make it incredibly difficult to stay polite.
This conversation is largely about the second group. And while I fully understand the worries brought up by GP, they've just about brought me to the point where I cease caring.
I'm Scandinavian. I would never have argued that it's okay to outcompete people. That isn't how our society works. But when a group of people spend this much effort telling me we're two different communities, and they don't care if I go off and hang myself? Well, they shouldn't be surprised if I accept their point of view, and decide they're a different community.
We spent so much time trying to put everyone in the same one, too.
Fortunately there's still the first group.
Again, I invite you to consider if that anger occurs in a context that might make it be or seem justified.
For example, many people are a little irate when they are on the cusp of losing nearly everything.
Many people now perceive they are on the cusp of losing everything. Are they, really? I'd say, yeah, probably.
Either way, your feelings about people who are angry at you are not as important as understanding those feelings, and understanding the feelings of the people who are angry at you.
Go for the understanding. "I'm fatigued by the anger" is not an appropriate response to a circumstance where your team may in fact have done something very wrong.
Are we the baddies? https://youtu.be/hn1VxaMEjRU
I meant: is our industry benefitting by eating other industries?
The answer seems to be "yes, obviously."
Does that make us the baddies?
Personally I'll add that I do not remember anything other than neutrality or more or less concealed glee from the same corners when the discussion was about how outsourcing and the like would eradicate my source of income, which arguably makes it much more to feel positive thoughts towards people advocating such proposals.
So, with regards to whether Filligree is with the baddies...
I recommend not invoking Godwin for puny things.
I want to know how far you take this 'permits hurting other people', because if me making a picture of Trump riding a unicycle counts then so should basically everything.
What you are groping towards is an attempt to find absurd counterexamples that throws doubt on the principle being proposed. This is fine, provided that your interlocutor didn't specifically allow for those counterexamples (which she did.)
If, as a general rule, your ethic permits you to ceteris-paribus harm others, you should probably at least examine your ethic.
That this sort of baseline humanism is now controversial speaks volumes about our cultural debasement.
We as an industry have brought immiseration as a primary product to the developed world for quite a while. Shoving human beings "below the API" (coined by Peter Reinhardt, but I heard it first from Venkatesh Rao) is what has gotten the richest in the tech space paid, and the idea that we can adjust our pince-nez and hmm-hmm at the people being actively harmed for not being sufficiently calm about the incipient transition from immiseration to active ruin is a silly idea when it isn't outright insulting.
It's not as horrible as it first tastes :D
I'm surprised that you hold Rawlsian self-respect in such high esteem but not Rawls's views on democratic institutions as key components to a fair society unless you think they're separable concepts.
Democratic institutions are key components to a fair society. They are deeply damaged. That contributes to the unfairness of our society.
Did I miss something?
Rawls developed a lot of his ideas as critiques of political philosophy in his time. One such school was the liberal tradition, works by authors such as Locke, Mills, and Kant, that argued that ensuring individual rights and happiness was the most important aspect of political society. Another school was the dialectical materialist school by authors such as Marx, Althusser, and Gramsci which argued that political society's goal should be to dissolve structural inequalities which arise from material conditions.
Rawls wanted to strike a middle ground between the two schools and acknowledge both materialist concerns and the need for individuals' rights. A key part of unifying these notions was to offer the idea of self-respect, that ensuring the key role of political institutions should be to maximize the self-respect of the individual not necessarily their rights or their autonomy. Self-respect is a lens that exists for social and political institutions to target, not an extant property of individuals themselves, they exist together [1]. The existence of state institutions is justified as tools to maximize self-respect and justice. As expected philosophers from both camps criticized Rawls's work, libertarians asserting that self-respect led to instances of infringements of rights and dialectical materialists claiming that Rawls's ideas would uphold bourgeoise inequality the same way democratic societies do now.
It's a problem I find with these sorts of discussions online. People use quotes from political philosophers to lend credence to their ideas without considering the totality of the philosopher's ideas. You can't reduce a philosopher into a pithy quote or tweet.
[1]: If you recognize Lacanian discourse, then you can view equity institutions as the Other with the messy bits of individuality and self-respect shaking out as the Real.
I'm well aware of Rawls' broader position, and I in fact hold an MA in phil with a concentration on Rawls.
Nothing you have said seems to interact with what I've said, other than to suggest, erroneously, that understanding was lacking.
Rawls does indeed suggest that "a key role of political institutions should be to maximize the self-respect of the individual", as you perfectly put it.
We are about to lose the sociopolitical institution that undergirds a huge amount of individual self-respect, namely, economic merit. If we had good sense, we'd stop right here.
(Anticipating the obvious: yes, the market for talent is a sociopolitical institution; if you need more on this, I'm happy to keep talking.)
Speaking of philosophical principles, I'm a big fan of 'the Principle of Charity' -- that one make an effort to parse text in a way that attributes as much wisdom and thoughtfulness to the interlocutor as possible. These days, it's often called 'giving a steelman' of your opponent (a riff on straw man, I would suppose.) https://en.wikipedia.org/wiki/Principle_of_charity
Regardless of what you prefer to call it: Might I suggest that you re-invest in the Principle of Charity? It would make our conversation more interesting. (This is its usual function.)
I want to emphasize that I feel that liberal democratic societies have deep obligations to ensure some economic stability for their citizens as well. (I draw a lot of inspiration from Rawls though I don't agree with him on everything.) I'm personally a fan of UBI here, but there are alternate proposals which ensure some form of economic dignity for members of liberal democratic societies. But shouting and raging and rejecting democratic institutions is not the way.
> We are about to lose the sociopolitical institution that undergirds a huge amount of individual self-respect, namely, economic merit. If we had good sense, we'd stop right here.
Forgive me in remarking that this sounds remarkably Burkean.
> Anticipating the obvious: yes, the market for talent is a sociopolitical institution; if you need more on this, I'm happy to keep talking.
Indeed, I'd love to see more on this. A Marxist might claim that the idea of economic merit and talent is but a facet of bourgeois society and a post-structuralist that the idea that one is owed economic merit at all is but a construction.
> Regardless of what you prefer to call it: Might I suggest that you re-invest in the Principle of Charity? It would make our conversation more interesting. (This is its usual function.)
Forgive me if I made assumptions, and I believe my tone came across as more combative than it was meant to be. But when you gesture at popular culture's disdain for techies as some form of proof of our sins, you make it difficult to take the argument seriously. Moreover my reply was meant more for whom I replied to rather than to you directly. You can accuse me of speaking "toward the audience" which I will admit is a bad habit of mine. But I've seen enough poorly understood "Paradox of Tolerance" and "Your fists' rights end where my face begins" shares on the internet that I don't always approach these conversations with the best of faith.
They weren't assumptions; they were careless reading. Don't do that again. And yes, popular culture -- art -- matters, and depictions (representations) that have currency also matter. If this makes it hard for you to take me seriously, as you say, then you seem to have entered this conversation with a view that culture is irrelevant to begin with, so why are we arguing about culture?
> Indeed, I'd love to see more on this. A Marxist might claim that the idea of economic merit and talent is but a facet of bourgeois society and a post-structuralist that the idea that one is owed economic merit at all is but a construction.
I'm no Marxist and I'm no post-structuralist. I have no opinion on what constitutes 'merit' in the absolute sense, but I observe that rapidly converting our current order to a new one, where cognition is cheap and provided by LLMs produced by large, central organizations, is probably not going to go well, mostly because, as a practical matter, people's self-respect is depends on their role in the economic order. Maybe that can change, but not easily, and not soon, and probably not in time.
> Forgive me in remarking that this sounds remarkably Burkean.
Again, loose reading. I am not claiming that actual merit (how much a human being is worth) has some connection to their economic output. I observe, as a matter of fact, that, in modern western democracies, we act as if this is the case; suddenly changing the rules of the game will have enormous consequences, many of them likely dangerous.
> You claim that shouting must happen because there's a conflict: the idea of individual self-respect (in one's economic value or merit) is eroding while liberal democratic institutions are incapable of dealing with change. This seems deeply contradictory.
I claim the shouting will happen because shouting is a natural human reaction to having one's sense of self-worth upended. And modern liberal democratic institutions are not likely to be able to handle this, yes, because they are sclerotic and polarized.
This is not contradictory; this is 2023.
And yes, a Leviathan moment is precisely what I fear might be coming: a liberal democratic capitalist order that fails over into some form of autocracy or another.
I hope I'm wrong.
To expand: no side is without sin in this. The art community has inclusiveness problems of its own, and you'll find that many of the loudest anti-AI voices are those privileged and connected enough to have been able to develop their craft and to make a living off it, not the majority of individuals who are struggling to find a place and a manner of expression that is successful (however they define it). The former's derisive calls to "pick up a pencil and draw, if you want to make art" are not altogether different from the advice you would get on deviantArt or some long-since-forsaken forum 20 years ago.
This, of course, is incomplete advice, especially in a quickly-developing art movement, where not having the resources, or the time, or the connections that produce skill or economic mutual aid, or the psychological space and safety to be self-critical and creative, could leave a would-be artist stagnated and frustrated.
Some people are able to find that capability, while some are merely left with years of resentment from trying and failing to break through. And, make no mistake: much of where one ends up has less to do with actual artistic ability, and more to do with one's social standing.
I imagine that many of the people who are engaged with AI art and its development fall into that latter category. Much as with crypto art and NFTs in 2021 and 2022, the wailing and gnashing of teeth from the anti-AI art crowd comes from the end-run people who were ignored or ostracized by those gatekeepers have been enabled to do around those "systems of respect" (which, as I've explained, were arbitrarily elitist and exclusionary to begin with). Their anger is self-serving, and in some regards cruel, because it's energy that a less egotistical group might have turned towards proving with action the morality of the process they espouse. But now, as before, the goal is not in enabling the success of those aggrieved who turn to AI as a balm, but in conserving an advantageous status quo. All the shouting does is cement this craven ethos as the true core of their discontent.
AI tools do redistribute the ability to produce art, which might otherwise tend to be limited to a lucky few who get the right education and training.
The problem, however, is that without a professional class, fields tend to drop in stature, and whole fields of human endeavour begin to seem less consequential. This is important to the self-respect argument I outlined above.
I recall an Onion article entitled something like "Area man asks friend if he knows how to get any doctoring work," about a man who (apparently) sees conducting surgery and doing odd-job home repairs as roughly equivalent.
Something like a "doctoring work" market is on its way for the artworld, and in the process, it will dramatically change the social significance of artists, the social stature of artists, and, more generally, the possiblity of making a defensible 'moat' around one's living as an artist.
The big beneficiaries of this process, aside from Susan, who will suddenly be able to write that novel, will be the guys selling Susan her novel, one token at a time.
Perhaps I'm old fashioned, but I genuinely think this will make Susan's novel less meaningful, both to her and to her readership, as well as making novels in general less meaningful.
All the while lining the pockets of, inter alia, my former employer.[1] I think this is.... not a good change?
[1] Disclosure: I benefitted financially from the advent of contemporary ML. It made my career, frankly. It stands poised to unmake the careers of many of my art friends.
Since Creative Commons is commonly used by charities, museums and volunteer organizations, they had some trouble maintaining the moral high ground. The attacks eventually fizzled out. (Even just finding back their texts attacking CC turns out to be pretty tricky)
But I still remember something of those attempts.
Just because some people are saying bad things about you, doesn't mean that those things are necessarily true.
What you're talking about is akin to the "professionalization" (or the maintenance of status) of a trade, artificially limiting the breadth of practice in order to direct energy to the quality. This is almost always portrayed as a good thing, when we've lived through an example of it (IT) being disastrous for workers. Professionalization puts up onerous barriers to entry; encourages cronyism; and strips a field of its base of novice workers, as well as a clear and germane path for advancement, replacing it with glad-handing and credentialism.
Collective action - unionization, foremost - would be a better solution in securing the livelihood of artists and conditions under which art is made, both for those who participate and even for those who don't. But we haven't seen it, or even the serious talk of it, and I imagine that it's for the reasons I've given: a central ethos rooted ego and not principle. And faced with a choice between shoring up the barriers for people who are only out for themselves, and opening the floodgates for expression, I choose the latter.
'Artist', as a thing that an adult can be and do, has only been around for a hundred or so years (said William Deresiewicz https://www.goodreads.com/book/show/53935644-the-death-of-th...) and there is no reason to think it will be around for a hundred more.
Yes, art has been around for thousands of years, but most of that time, it was beholden to: the church, the power structures of the day, etc.
The idea of 'art for art's sake', the idea of artists empowered to do and make things that their commission did not specifically tell them to do, is not an immortal category of human.
I'm not doing this argument justice, please read William Deresiewicz. <3 What a great book.
Current AI output looks impressive, but if you push hard you quickly start hitting limits, and you discover you're going to need to hire/cooperate with experts in several fields if you want to take things further.
In future, rather than single artworks, people are more often going to want LoRAs, checkpoints, models, etc. And it already seems like they are willing to pay for these things to be made. There are definitely opportunities for new business models to be had here.
> "There are systems of respect at stake..."
I feel like you're overextending a particular sociological factor beyond all reason. Human beings draw meaning from a great many things, setting meaningful contribution as a necessary factor needs way more justification than just quoting one political philosopher.
You only need to look at our current society to cast doubt on this assertion. We live in capitalist system that promotes individualism more than anything else. Anyone working in a job right now is acutely aware of the fact that they can be replaced - not with AI but with a different human worker - at any point. We don't live in small villages anymore. Every baker in a city has to cope with the reality that his customers would, without a care, switch to the bakery across the street if he were to close down. Not only that, but many people "contribute" by doing work that is neutral for society at best and actively harmful at worst, as you yourself pointed out.
Yet our society still stands, and has stood for quite some time. People can adjust, they can derive meaning from sources that have been neglected or ignored, and live life just as happily as they did before.
> "A quick closing word..."
I think you're being overly defeatist here. There is a reasonable chance that giant challenges are ahead, but the extent of these challenges is still unknown. There are a great many possible scenarios where just a little constructive discussion and some social policies may avert most of the negative consequences. The more we do now, the more likely it is that we will persevere in the coming storm and emerge with minimal damage.
I think the industrial revolution is a good comparison, even though I believe that the AI revolution is going to be far more challenging. The industrial revolution caused quite a bit of suffering in the short term, but you and I probably agree that this suffering could have been strongly minimized with the right approach. We can learn from the lessons provided to us from history and sociology, and give it our best shot.
And as for political inertia, we coasted on the "good enough" setting for quite some time now. Looming threats, especially ones where consequences to the individual are easy to understand, can fuel quite a lot of political action.
Suggesting that there is nothing that can be done is usually easy - after all, it absolves us of any need to actually do something - but it is also the best and fastest path towards the worst-case scenario.
EDIT:
I'm not suggesting that it's possible to have a discussion that is completely detached from emotion or the existential dread. The issue that I have is that we are stuck having discussions that are completely orthogonal to the issue at hand. I see pages and pages of heated arguments about copyright, the nature of art and other such trivialities. Discussions about AI inevitably deteriorate into an emotional circlejerk, where showing up the other side is the main motivation. None of that is going to help us deal with the fallout of what's coming.
The caricature here is entirely yours. Your interlocutors aren't raging against evil tech workers, they are railing against tech workers who are studiously blind to their own industry's externalities.
You do the thing for which you blame your adversary.
> I feel like you're overextending a particular sociological factor beyond all reason. Human beings draw meaning from a great many things, setting meaningful contribution as a necessary factor needs way more justification than just quoting one political philosopher.
Sir, this is not a textbook, this is the comments section on a popular Internet website. If you want 'way more justification', you will have to wait for my thesis.
Tech workers - like most people - are unaware of the extent of suffering that they don't directly perceive. You and I know that there are people suffering in third-world countries - but knowing is not understanding or awareness, and if we had to spend some time in the company of these people, our perspective would shift enormously and we would feel a much stronger urge to actually remedy these injustices.
Tech workers know that they are displacing workers, but again, knowing is not understanding. Hurling insults and making sly accusations is not going to convince anyone that they may lack perspective on this issue. I mean, take a look around this thread, take a look around other discussions - I don't see many people trying to convey their experience without painting the tech crowd as money hungry assholes.
Not to forget that AI isn't evil either. You can make many moral arguments for developing AI systems as fast as possible. The most emotionally effective point would be medical research - after all, who are we to place our own lives above the ones that may be saved with the help of future AI developments?
And another note: many tech workers argue that it is irrelevant whether or not they personally continue to develop AI, because someone most definitely will.
Also, there is no adversary. I would love for the AI skeptic crowd to make compelling and convincing arguments. We need people that warn about the problems introduced by AI and people that campaign for policies that address them, but we need them to do it in a way that would actually be able to change someone's mind. The reality is that we're not having the right discussions in a productive way; instead we pontificate endlessly on irrelevant but emotionally satisfying details.
> Sir, this is not a textbook, this is the comments section on a popular Internet website. If you want 'way more justification', you will have to wait for my thesis.
Do you really think there is no midpoint between making a massive, fundamental assertion about the stability of every human society without any further elaboration or studies and writing a textbook or thesis?
Pretty standard human behaviour, I'd say. Reference points include Exxon, Shell, et al, at any time in the past forty years. The crucial part is that it's possible to be aware at one level and deny the problem thoroughly at another. Please do not tell me I have to prove the existence of the possibility of the psychology of denial? :)
> Because the word "studiously" implies that they are aware of what they are doing.
That is exactly what they are doing. They are constructing themselves as unaware. As Upton Sinclair put it, it's hard to get a man to understand something when his paycheque depends on not understanding it.
> You yourself said that tech workers, or "tech bros" as you so nicely put it, are perceived and portrayed as uncaringly and knowingly destroying peoples life.
I didn't say uncaringly, but yes, for certain values of 'unknowingly', this is the case. Contrived innocence is the order of the day.
One example (there are many): Tech bros in major west coast cities with large homeless populations routinely pass within centimetres of people whose suffering they -- through capitalist competition -- underwrite, whether they want to or not. We have priced whole communities out of their historic neighbourhoods. Evictions are commonplace, and despite the fact that we know they help manufacture unnecessary poverty, they are allowed to persist, in part because we -- the housed professionals -- are the quiet beneficiaries. (See recent HN-trended https://www.newyorker.com/magazine/2023/03/20/matthew-desmon...)
The main psychological tools I see deployed to deal with the so-called "unfortunate" (take a moment to unpack that word) are: aversion, bad faith (in the Sartrean sense,) angry denial (anger is such good anesthetic) and libertarian dogma that falsely localizes the problem either in inevitable facts-of-life or personal failure.
We are now extending that 'read' to include those who are about to be professionally (and perhaps literally) displaced by the GPT revolution.
> If you think that that is a fair assessment of most tech workers motivations, then you are constructing a caricature.
See, again, you managed to construct a caricature of the position of the person telling you that you'd constructed a caricature.
This is impressive. Would suggest introspection.
You understand that by "you" I was referring to people who have that perspective right, as highlighted by the fact that I wrote "are perceived and portrayed" before, not "are perceived and portrayed by you"? Should I now tell you that you too managed to construct a caricature of the position telling you that you constructed a caricature? Is that where this series of le epic owns leads? I'll make sure not to include any argument as to why I think it's a caricature either.
Maybe I should add some snarky remark as well, now that'd be a good use of both of our time, right?
We are, however, talking through each other.
I'm directly addressing the stance that you have repeatedly taken. I quote:
> When someone is aware of the damage they cause, and then they still continue, what would you call that behavior? Because the word "studiously" implies that they are aware of what they are doing. You yourself said that tech workers, or "tech bros" as you so nicely put it, are perceived and portrayed as uncaringly and knowingly destroying peoples life. If you think that that is a fair assessment of most tech workers motivations, then you are constructing a caricature.
I don't know how to read 'If you think that this is a fair assessment of most tech workers motivations" as anything other than a claim that I think you think you mean that I think that this is a fair assessment of most tech workers motivations. You seem to be chasing a 'gotcha' moment, where I have to either admit that tech workers know full-well that they are doing something wrong (and are evil, because they keep doing it) or did some bad things, but didn't really know better (those folks in developing countries that we could get to know better but don't.) This seems to miss, on purpose, the possibility of knowing something at one level, and denying it at the other, a basic feature of human psychology that (I took) us to all have understood as a possibility when we started talking.
One can be judged for what one refuses to understand. For the understandings that one systemically avoids.
> Tech workers - like most people - are unaware of the extent of suffering that they don't directly perceive. You and I know that there are people suffering in third-world countries - but knowing is not understanding or awareness, and if we had to spend some time in the company of these people, our perspective would shift enormously and we would feel a much stronger urge to actually remedy these injustices.
First, 'third world' is not a phrase you should be using in 2023. It implies rank ordering, and no sociologist would have touched it in the past forty years. Please tweak your language.
> Secondly, Tech workers know that they are displacing workers, but again, knowing is not understanding.
Avoiding understanding is precisely what we're talking about.
> Hurling insults and making sly accusations is not going to convince anyone that they may lack perspective on this issue. I mean, take a look around this thread, take a look around other discussions - I don't see many people trying to convey their experience without painting the tech crowd as money hungry assholes.
No one is hurling anything. I am calmly answering the points you raise as I take you to have raised them. If I have mischaracterized your position, I am genuinely sorry, and it was not intentional.
But it sure sounds to me like you want to absolve some folks that I don't think qualify for absolution. We may simply disagree as to what falls within one's personal moral ambit. This is a conversation we can have at length. I'm here for it. <3
On a personal level, I'm not happy to see a medallion-owning taxi driver lose their livelihood, in the same sense that I'm not happy to see anyone lose their livelihood, e.g., due to economic conditions, due to a company shutting down, or any other reason.
At the same time, there will be winners and losers from any economic or technological shift, and it seems that in our capitalist society we value progress and competition, underpinned by some utilitarian calculus that the short-term thrash is worth it for the long-term prosperity.
Inventing the telephone certainly would have harmed mail couriers, inventing the automobile would have harmed horse breeders, etc., but we don't typically reflect on those inventions as being ethically problematic.
In short, what I'm wondering is whether you think there's something uniquely unethical about the "tech bro" style of disruption, or if your perspective is more broadly a critique of capitalism.
The distinguishing characteristic is scale & speed.
The switch to electricity happened over half a century. The switch to the automobile took a similarly long time. There was crossfade. There was time to adjust. There were places (for socioeconomic values of 'places') for skilled professionals to go.
The switch to Uber (from taxi medallions) seems well underway at just the one-decade mark.
Differences in speed & scale yield differences in qualitative impact, in the same way that a nuclear bomb -- which increases speed & scale of destruction, over and against TNT -- produced a whole new circumstance for world powers.
Comparing what is coming to telephony replacing a courier is like comparing Nagasaki to dynamiteing an old tree stump on a farm.
Because: What this does is, it makes it harder to train models on specific works. Okay. But if the model is sufficiently generalized, an arbitrary style could simply be described to the model (not by the artists name, but by technically describing the style), and then reproduced, even if the model never saw any training data in that style.
Come to think of it: If a model is sufficiently generalized, it could reproduce any conceivable style described to it, not just those it didn't have in its training set, but even styles that didn't exist before.
So yeah, mid-to-long-term, how does this "protect from style mimicry"?
Further, if we're going to accept that style mimicry is enough to categorize something as being plagiarized, every artist on the planet will have to be labeled as a thief as well
Some nations already feature legislation close to that, sans the standardization of the format.
'Just' getting this through the EU, the US and China would probably be impossible and even then there would be plenty of jurisdictions left over.
It has a chance of working.
I can't see that chance for attempts to dissuade people by manipulating images. Because the latter can lead to an arms race: Training workflows get smarter and reverse or otherwise deal with changed data points. In response, technologies changing the data to prevent training will have to get better as well.
If the ingestion workflow alters the picture sufficiently, the protection could be lost, same as DRM qualities are lost if someone alters the data-stream by recording what is shown on the screen of the display device.
[It should be noted that the comment 3 up the chain i was replying to got edited so my response 2 up the chain makes less sense now]
There have even been other drm systems that have fairly succesful at attacking analog hole, e.g. https://en.wikipedia.org/wiki/Cinavia
Of course, no solution prevents everything,but preventing a specific system from using the data in a way that is hard to get around is highly possible.
OTOH, that’s also one of the problems with it for art that is intended to be displayed to humans (they minimize this descriptively, but the full-size examples I’ve seen are bad.)
In this case, you don't need to do anthing special to work around this tool, just do what everyone is already doing and resize all your training images to a common size, thus defeating Glaze entirely.
Regardless of your stance on the AI art/IP debate, it seems enormously hypocritical for AI builders to say that they not only have the moral right to use images against the artists’ wishes, but that the artists also have a moral obligation to provide these AI builders with clear, denoised versions of their work.
Edit: if you use a tool to thwart internet data collection systems, thus making your data less useful for advertisers, are the data collectors justified in bitterly calling you a “malicious user”?
If I asked how artists can thwart “malicious” tools like Midjourney, you could reasonably assume that I disapprove of Midjourney and think they ought to stop doing what they do.
What are the best ways for crackers to defend against malicious tools like copy protection?
It's because scores of people coming from non-art domains are suddenly being exposed to the art domain without much real knowledge of working in that domain, find the diffusion tech convincing enough that it entrances them, and makes them believe that their original notions of art are the most important thing that matters.
This debate has made me think, if there's some inability to automate a process, only the people that have spent months/years working in those domains will understand certain things about that process. The rest will be operating on only the beliefs they form from limited exposure. For example, I imagine for a lot of people, "art" is something that magically appears in their Twitter timeline within 5 seconds of opening the app, and years of opening Twitter and seeing new art almost instantly would probably condition some to believe that's all art really is, a thing that's plentiful and can be found near-instantly. So it's not controversial to them when they come across a piece of software that can give them thousands of new pictures near-instantly.
But obviously, prior to around October of 2022, there was nearly always a human behind the art on those timelines. To a human artist, their own work is probably not as plentiful and takes way longer to produce than to consume. Those consumers never needed to interact with the artist or understand the process to appreciate the art privately. Up to now that was usually not a problem.
Maybe that idea will undergo changes in the following years, that the tech is so advanced that a level of respect for the specific process of human artists is necessary for cordial communication of art to succeed. Otherwise it will be disparaged with the equivalent of "opinion discarded" from the nature of the thing.
It makes me wonder, what are the implications of this for domains that haven't been automated to the extent of SD yet? SD has proven that many, many people wanted this level of "art" automation from the beginning. The singular event that made this fact clear was the release of SD. There could be millions of people out there with their own preconceived notions, correct or not, just waiting for the ability to cut out the middleman in their own domains. It's like a powder keg waiting for the tiniest spark of technological progress to be ignited.
For the "respect of artists" angle, I'm reminded of this article[1] about a modern art installation from 12 years ago where people's art contest entries were trampled upon by visitors. People were upset by this.
As Marshall McLuhan once stated, "art is anything you can get away with."
Of course, if you are a mind-body dualist or support idealism, you will think that humans are somehow different than other arrangements of matter that give rise to computation.
As someone who does creative work (writing and animation mostly), posts it publicly for free on the Internet, and tries to actively help people learn (by, e.g, answering questions when somebody asked how I did something or what my thought process was), I'm very much in favor of humans learning from something I made, and I'm also very much against a model like ChatGPT or Stable Diffusion learning from something I made. Let me explain my stance:
There's a lot of reasons to help humans that don't exist for machines. The biggest one (and I suspect this to be the case for a lot of artists) is that anybody that's trying to learn an artistic endeavor is automatically somebody that I relate to. Learning to do art well is hard. I think any good artist is in a constant state of struggling to improve, so they're very going to be very much cognizant of how it feels to struggle to improve, and it's highly likely (or at least my perception is that it's highly likely) that anybody seriously studying somebody else's art and trying to learn from it is experiencing something similar, and has similar goals and motivations. In that situation, it's only human to want to help the other person out.
That doesn't apply to a machine. Stable Diffusion is not living with three roommates with no health insurance and only one pair of pants with no holes in them, earning rent and food off spotty freelance work for random people on Twitter. ChatGPT isn't burning the candle at both ends working a day job and spending long nights trying to piece together a short story good enough to get published in an open call fiction magazine with a 1.1% acceptance rate that pays ten cents a word. There is no toiling human making sacrifices because they were born with an irrepressible urge to create; instead, it's arguable that the primary reason these things exist is to provide some return on investment for the huge companies that own them. Microsoft's dumped billions of dollars into OpenAI. It must be expecting to make that money back somehow, so helping ChatGPT out feels like I'm just making some faceless suit even richer than they already are, which is sort of the exact opposite of helping out a fellow creator.
(There's an argument to be made that things like ChatGPT are providing a net positive to society, so I should feel happy to help train them on that count, but in their current state I don't think they are. That's a separate ramble though, and a tangent I don't really feel like getting into while I'm explaining my stance here).
The outcome I expect from helping a human is also different from the outcome I expect from helping a machine. Humans are different from machines. For a human to copy somebody's style, I think they have to fully understand the process that the other human took to arrive at that result. To do that takes solid fundamentals, and to develop those fundamentals you have to spend a lot of time practicing, which means exposing yourself to new influences. I really do think it's pretty much impossible for somebody to completely copy somebody else's style. When you create something, it's a deeply personal experience. What resonates with you? What image, or theme, or emotion, or experience, spoke to you on such a deep level that it drove you to spend forty minutes trying to animate a foot so it slaps against the ground in just the right way? Is it really likely that any other individual is going to see what you made, and get out exactly what you put into it, without putting any of their own spin on it? In my experience, it doesn't happen at all. So when somebody copies you (as long as the copy was effortful), it's a joyous experience to see what they made. It's interesting and illuminating and fun to see an earnest expression of how somebody else sees your art, to the point where people do stuff like this: https://static.wikia.nocookie.net/85799efa-2901-4cf7-a05d-a5... (I don't have any special connection to WoF, it just happened to pop up when I googled 'draw other artist style meme')
OK, but somebody with solid enough fundamentals could get /close enough/ that it doesn't matter, and then sell to less discerning members of your audience, right?
Well, I guess in-principle that's possible, but it just really doesn't happen. Anybody that had the drive to get to the point where they could copy somebody else and have it look good probably already has their own style they're proud of. I don't know a single artist that would be happy with a client coming to them and saying "hey can you draw me something and make it look exactly like <other artist> drew it?". In the worst case they'd tell you to fuck off, in the best case they'd begrudgingly accept your money but probably be insulted and not put as much effort as they would be working on something personal and exciting for them. So another person learning from you isn't infringing on your audience and livelihood. Or if it does, the impact is so small that the positive feeling of helping someone similar to you out outweighs that by quite a bit.
Machines aren't like that. Machines will gladly slurp up your entire body of work, churn through it in training, and then shamelessly crap out a limitless supply of stuff that's rehashes of things you made for somebody that wrote 'by <your name>' in the prompt. You don't even get the warm fuzzies of knowing you helped somebody out, or the cool experience of seeing somebody else's interpretation of your art. It's just an uncanny Frankenstein of your work that's a sordid reminder that a good portion of people out there don't see artists as people, don't engage with art the way you do, and don't want to respect your wishes that the stuff you make not contribute to a cash-and-power shift that's decidedly not in your favor.
What happened is that AI builders took artists’ images, baked them in GPUs for a few months, and are now selling the output. Again, regardless of your opinion of the IP question, you have to admit that generative AI leverages artists’ own work against their economic interests. If someone’s livelihood is affected by this, of course they will feel venom. That is not parallel to your feeling of venom toward their efforts to stop this from happening to them.
They're not actually supposed to have anyone else's pictures in them as such [1].
As far as I can tell, the training is more along the lines of "this is what a dog looks like", "this is what a cat looks like", "this is what a stick looks like" ... etc.
At this point in time, unaltered output is automatically public domain, because it is not made by a human.
[1] some early bugs notwithstanding.
Yep, this is exactly the kind of take I'd expect from the AI crowd.
The obvious conclusion is that I'm not tricked into thinking that some one is somehow deprived just because you bought into the 'piracy' propaganda. Side note to consider; why is it called piracy? Would many of the arguments offered by people trying to maintain their dragons horde of IP and the monies from that IP be weakened if we called it what it was - downloading a copy?
Most people are like this - they will gladly buy the things they enjoy when they can afford to. The scale gets tipped when the cracked version works better than the official version, or when some one lacks the disposable income. Its just really hard to demonstrate harm here because there is no harm.
Or just... downvote, lol. best argument ever, HN intellectual debate on full display.
The tool changes the image. Okay. So does cropping, changing color composition and a gazillion other things.
Wanna clearly communicate that intent? robots.txt. Exif data. Attributes in `<img>` elements.
Technically? Incorporate Glaze into the training process.
Glaze takes advantage of an ML model's fragility towards noise and uses that against the model.
The simplest way of incorporation is to run Glaze on all of the training data, whilst duplicating
Glaze, reductively speaking, is the adversarial part of a GAN, wherein MidJourney/SD is the generator part of the whole system. What's happening now is that the adversarial part of the GAN is purposefully producing noise to throw the generator off of its rocking horse, and now the generator needs to learn to "remove" said noise when training.