3 ms·

As the saying goes, you can't put pandora back in the bag. AI art is here to stay, and it's here to dominate. The problem is the economic system within which ar

As the saying goes, you can't put pandora back in the bag. AI art is here to stay, and it's here to dominate. The problem is the economic system within which art is created and assigned value, any attempts to solve the "problem" of AI art without addressing the underlying economic incentives are destined to fail in the same way that we have failed in preventing the use of sweatshops by clothing suppliers. Even if we have strong, even overbearing legislation put into place to protect artists, we're just going to end up with the biggest most profitable offenders buying/using illegally trained models through middle men, and feigning ignorance when/if they are found out. Though most of the "has this model been trained on X" conundrum is likely to be irrelevant on a years (months?) timescale as art styles are far less unique than artists would like to think. See tencent's PhotoMaker or other modified CLIP approaches. Even a model not trained on a particular face can be conditioned to generate said face using a oneshot approach because the vector representing that face can be constructed even if the face itself doesn't exist in the training data. I'm certain the same is true for artistic styles.

throwaway290 2y ago
Conveniently defeatist. "AI" tech? yes we can, scrape the entire web, fire up the transformers (aka roll that VC money)! "prevent stealing from creators" tech? nah, too difficult, we don't know how, human problem (the neighborhood queer artist can't pay nearly as much). Fingerprinting technologies. Generative tools can be mandated to embed watermarks. Original images can include fingerprinting and other counter-theft measures that tools must respect. Offenders who publish tools that don't can be prosecuted a la tornado cash. etc It's not about art style. The model exists thanks to data, period. No data, no vectors. And almost all of that data is stolen.
- thot_experiment 2y ago
  Tornado Cash was a service that people used, Stable Diffusion is a 2gb blob of tensors. The latter is not something you can control, the MPAA/RIAA have been trying to control the proliferation of pirated media for 30 years, it has not worked and their case is WAY more clear cut under our current system of laws. Additionally, there's no way you will be able to force people to insert fingerprinting into the model or surrounding tooling. There is no way to force fingerprinting into the source material, glaze etc are memes that are easily defeated and in fact can help stabilize outputs when training on glazed images is used as a negative embedding. The idea that any of this stuff is a viable method of combating gen AI is laughable. You can run this stuff on an 8 year old GPU with a blob the size of a couple netflix episodes. While outputs straight from SD may be identifiable, SD + manual intervention can easily create outputs that are not traceable. There is no workable solution that doesn't address the underlying economic incentives. It doesn't matter that the model is trained on data that you believe is "stolen", people who want the model will have access to it, you won't be able to prove the outputs are AI, and even if you could, it won't be cost effective to prosecute people for creating in this way. The problem here is our economic system and it's reliance on scarcity and markets to assign value to things that are fundamentally not scarce (AI models, digital goods) and it's failure to correctly assign value to things that have enormous positive externalities (artists creating culture).
  - throwaway290 2y ago
    > there's no way you will be able to force people You'd be surprised. Pass a law requiring ClosedAI to start doing it tomorrow and what will they do?
    - thot_experiment 2y ago
      This is not how that works my brother. The AI is already on my computer and the computers of millions of other people, it doesn't need to talk to the internet and uses an entirely open source stack. The minimum system requirements are a gaming computer from the last decade.
      - throwaway290 2y ago
        Nah mate you'll need to continuiusly acquire new data to stay updated, otherwise tomorrow a genius artist comes up and if your model can't imitate her style, what a sad day? If you have the resources to scrape the entire web and sanitize the data and fight the anti-AI arms race and retrain the model all the time and you do it alone just for yourself at home for the lulz then I guess knock yourself out, no one will mind. But a gaming computer from last decade alone probably won't cut it > entirely open source stack I heard of some tornado cash and wow using OSS for crime totally made it not a crime! > millions of other people Yeah sure my phone also has ML
        thot_experiment 2y ago
        Bruh, the idea that tornado cash and SD are somehow equivalent is laughable. The point is that the open source stack means that there's no point where you can force any of these fingerprinting things in. It's all already out in the open. It's just like HDCP, if I can see the video on my computer I can pirate it, the same goes for glaze and related approaches, fundamentally you are not going to be able to prevent someone from training on an image if a human can correctly interpret the image as the artist intended. Sure you may cause some extra steps, but those steps will quickly be automated away. The vast majority of the "theft" from artists in the form of LoRAs and finetunes is done by individuals and trivially propagated all over the internet. I can train a LoRA on a particular artist in about 6 hours using my old ass computer and a base model that's almost 2 years old plus 20 images scraped from wherever and achieve great results. Most of what gives the model the generic capability to create images doesn't need to be retrained, finetuning on styles is very cost effective. Though many of the advances in the capability of SD have nothing to do with training the model at all but rather advances in instrumentation and conditioning. Art is like 90% building on generic work that others have done before you and like 10% your own contribution if that. Everything is a remix. That is the the core truth that allows style transfer approaches to work even if the specific style is not represented in the training data, in most cases there's a blend of conditioning vectors that gets you pretty close. Maybe the future of art is "making art that's hard for the computer to recreate" which would be dope, but the stuff it's already good at? That's over and done with, that cat is not going back in the bag. And to be clear, the set of stuff it's good at is larger than the set of things that are in the training data because of advanced conditioning techniques like CoAdapters/IPAdapter/ControlNet/modified CLIP conditioning etc. And yes, millions of people, SD 1.5 had 4.5mil downloads in the last month alone.

throwaway290 2y ago

Conveniently defeatist. "AI" tech? yes we can, scrape the entire web, fire up the transformers (aka roll that VC money)! "prevent stealing from creators" tech? nah, too difficult, we don't know how, human problem (the neighborhood queer artist can't pay nearly as much). Fingerprinting technologies. Generative tools can be mandated to embed watermarks. Original images can include fingerprinting and other counter-theft measures that tools must respect. Offenders who publish tools that don't can be prosecuted a la tornado cash. etc It's not about art style. The model exists thanks to data, period. No data, no vectors. And almost all of that data is stolen.

thot_experiment 2y ago
Tornado Cash was a service that people used, Stable Diffusion is a 2gb blob of tensors. The latter is not something you can control, the MPAA/RIAA have been trying to control the proliferation of pirated media for 30 years, it has not worked and their case is WAY more clear cut under our current system of laws. Additionally, there's no way you will be able to force people to insert fingerprinting into the model or surrounding tooling. There is no way to force fingerprinting into the source material, glaze etc are memes that are easily defeated and in fact can help stabilize outputs when training on glazed images is used as a negative embedding. The idea that any of this stuff is a viable method of combating gen AI is laughable. You can run this stuff on an 8 year old GPU with a blob the size of a couple netflix episodes. While outputs straight from SD may be identifiable, SD + manual intervention can easily create outputs that are not traceable. There is no workable solution that doesn't address the underlying economic incentives. It doesn't matter that the model is trained on data that you believe is "stolen", people who want the model will have access to it, you won't be able to prove the outputs are AI, and even if you could, it won't be cost effective to prosecute people for creating in this way. The problem here is our economic system and it's reliance on scarcity and markets to assign value to things that are fundamentally not scarce (AI models, digital goods) and it's failure to correctly assign value to things that have enormous positive externalities (artists creating culture).
- throwaway290 2y ago
  > there's no way you will be able to force people You'd be surprised. Pass a law requiring ClosedAI to start doing it tomorrow and what will they do?
  - thot_experiment 2y ago
    This is not how that works my brother. The AI is already on my computer and the computers of millions of other people, it doesn't need to talk to the internet and uses an entirely open source stack. The minimum system requirements are a gaming computer from the last decade.
    - throwaway290 2y ago
      Nah mate you'll need to continuiusly acquire new data to stay updated, otherwise tomorrow a genius artist comes up and if your model can't imitate her style, what a sad day? If you have the resources to scrape the entire web and sanitize the data and fight the anti-AI arms race and retrain the model all the time and you do it alone just for yourself at home for the lulz then I guess knock yourself out, no one will mind. But a gaming computer from last decade alone probably won't cut it > entirely open source stack I heard of some tornado cash and wow using OSS for crime totally made it not a crime! > millions of other people Yeah sure my phone also has ML
      - thot_experiment 2y ago
        Bruh, the idea that tornado cash and SD are somehow equivalent is laughable. The point is that the open source stack means that there's no point where you can force any of these fingerprinting things in. It's all already out in the open. It's just like HDCP, if I can see the video on my computer I can pirate it, the same goes for glaze and related approaches, fundamentally you are not going to be able to prevent someone from training on an image if a human can correctly interpret the image as the artist intended. Sure you may cause some extra steps, but those steps will quickly be automated away. The vast majority of the "theft" from artists in the form of LoRAs and finetunes is done by individuals and trivially propagated all over the internet. I can train a LoRA on a particular artist in about 6 hours using my old ass computer and a base model that's almost 2 years old plus 20 images scraped from wherever and achieve great results. Most of what gives the model the generic capability to create images doesn't need to be retrained, finetuning on styles is very cost effective. Though many of the advances in the capability of SD have nothing to do with training the model at all but rather advances in instrumentation and conditioning. Art is like 90% building on generic work that others have done before you and like 10% your own contribution if that. Everything is a remix. That is the the core truth that allows style transfer approaches to work even if the specific style is not represented in the training data, in most cases there's a blend of conditioning vectors that gets you pretty close. Maybe the future of art is "making art that's hard for the computer to recreate" which would be dope, but the stuff it's already good at? That's over and done with, that cat is not going back in the bag. And to be clear, the set of stuff it's good at is larger than the set of things that are in the training data because of advanced conditioning techniques like CoAdapters/IPAdapter/ControlNet/modified CLIP conditioning etc. And yes, millions of people, SD 1.5 had 4.5mil downloads in the last month alone.