5 ms·

Well, that was the entire point of banning AI generated content, right? Or, are you implying that the article states different reasons for the ban?

by XCSme 3y ago

freedomben 3y ago
> that was the entire point of banning AI generated content, right? But AI generated content is NOT banned. You just have to prove you have the copyright (or permission) for the training data.
- mrweasel 3y ago
  Which honestly sounds rather reasonable and should be the expectation of all current AI products. The only reason it wouldn't be easy enough to provide is if you just scraped any available data set with a complete disregard for intellectual property.
  - Matticus_Rex 3y ago
    Humans take in inputs and transform what they learn into distinct outputs. Training data is essentially just a machine doing the same thing. Knowingly scraping pirated material would be one thing, but essentially having the machine view publicly-available material is not clearly at odds with existing IP law.
    - mrweasel 3y ago
      And if your output isn't distinct enough from the inputs, you too aren't allowed to claim copyright or sell your work without proper licensing. With the AI we can at least be 100% certain of which input you trained it on and under which licens, making the whole a lot easier to deal with, as compared to humans. The liability is the same, but it's much easier to avoid legal implications, so why not play ball and ensure that you have the correct licenses?
      - Matticus_Rex 3y ago
        Have what correct licenses? It's not clear that you need any license to train on data in the vast majority of cases. Having a license to train on it won't guarantee that you can grant your users a license to any particular output, especially given the addition of user input. And most of the utility is in creating outputs that are indeed distinct. So the answer to "why not play ball?" is: 1. It's not clear that it's legally required 2. It would be incredibly expensive and/or slow progress dramatically, or limit you to pre-existing licensed content (e.g. Adobe) which drastically reduces some types of capabilities 3. Given #1, for any company that doesn't have an Adobe-style library, "playing ball" is essentially betting the company that it will become legally required, because on top of developing an AI model you're going to have to become an expert content licensing and documentation studio
    - __loam 3y ago
      What humans do when they learn is a completely separate issue and has absolutely nothing to do with how this technology is regulated.
      - Matticus_Rex 3y ago
        It's the most analogous process we have; the courts may not end up treating them the same, but it's unlikely they'll say it has nothing to do with how AI gets regulated.
nickelcitymario 3y ago
I may be wrong, but I believe they were just intending to summarize.
kemayo 3y ago
I'm quoting the relevant bit from Valve's email in the reddit post, to indicate exactly what they're actually forbidding. Notably: not all AI-generated content, but rather AI-generated content from models that were trained on material that's not owned by the person submitting the game.
- XCSme 3y ago
  Can that even be proven/tested?
  - BlueTemplar 3y ago
    Depending on the court, burdens of proof can vary a lot...
    - dragonwriter 3y ago
      In court, it would be the entity claiming infringement that would have the burden of proof that their exclusive rights under copyright were violated by the assets in question, not the distributor of the asset that had the burden of showing ownership of every item in the training set of the model used in some part of the workflow.
      - kemayo 3y ago
        What I'd assume Valve is worried about is that it only takes one major decision against Stable Diffusion in court to suddenly leave us in a state where "this game used Stable Diffusion" is the proof that's needed. Given the whole "Stable Diffusion reproduces the Getty Images watermark" lawsuit[1] that's still ongoing, it's not an idle concern. [1]: https://www.theverge.com/2023/1/17/23558516/ai-art-copyright-stable-diffusion-getty-images-lawsuit https://www.theverge.com/2023/1/17/23558516/ai-art-copyright...
        dragonwriter 3y ago
        > What I'd assume Valve is worried about is that it only takes one major decision against Stable Diffusion in court to suddenly leave us in a state where "this game used Stable Diffusion" is the proof that's needed. Hard to see how any plausible outcome that would have that result for users of SD (if model training isn’t fair use, that’s definitely a blanket-liability issue for Stability.AI — and Midjourney, and OpenAI, and lots of people training their own models, either from scratch or fine-tuning, using others' copyright-protected works. But “using a tool that violates copyright in the workflow” is not itself infringement; whether and in what situations prompting SD to produce output makes the output a violation of copyright (and whose) would be a completely different decision, and while Ibcan certainly see cases (such as deliberately seeking to reproduce a particulaflr copyright-protected element, like a character, from the source data) where it might be (irrespective of the copyright status of the model itself), I haven't seen anyone propose a rule that could be applied (much less an argument that would justify it as likely) based on copyright law that gets you to “used SD, in violation”. Lots of blanket ethical arguments about using it, but that’s a different domain than law.

Well, that was the entire point of banning AI generated content, right? Or, are you implying that the article states different reasons for the ban?

by XCSme 3y ago

Well, that was the entire point of banning AI generated content, right? Or, are you implying that the article states different reasons for the ban?

freedomben 3y ago

> that was the entire point of banning AI generated content, right? But AI generated content is NOT banned. You just have to prove you have the copyright (or permission) for the training data.

mrweasel 3y ago
Which honestly sounds rather reasonable and should be the expectation of all current AI products. The only reason it wouldn't be easy enough to provide is if you just scraped any available data set with a complete disregard for intellectual property.
- Matticus_Rex 3y ago
  Humans take in inputs and transform what they learn into distinct outputs. Training data is essentially just a machine doing the same thing. Knowingly scraping pirated material would be one thing, but essentially having the machine view publicly-available material is not clearly at odds with existing IP law.
  - mrweasel 3y ago
    And if your output isn't distinct enough from the inputs, you too aren't allowed to claim copyright or sell your work without proper licensing. With the AI we can at least be 100% certain of which input you trained it on and under which licens, making the whole a lot easier to deal with, as compared to humans. The liability is the same, but it's much easier to avoid legal implications, so why not play ball and ensure that you have the correct licenses?
    - Matticus_Rex 3y ago
      Have what correct licenses? It's not clear that you need any license to train on data in the vast majority of cases. Having a license to train on it won't guarantee that you can grant your users a license to any particular output, especially given the addition of user input. And most of the utility is in creating outputs that are indeed distinct. So the answer to "why not play ball?" is: 1. It's not clear that it's legally required 2. It would be incredibly expensive and/or slow progress dramatically, or limit you to pre-existing licensed content (e.g. Adobe) which drastically reduces some types of capabilities 3. Given #1, for any company that doesn't have an Adobe-style library, "playing ball" is essentially betting the company that it will become legally required, because on top of developing an AI model you're going to have to become an expert content licensing and documentation studio
  - __loam 3y ago
    What humans do when they learn is a completely separate issue and has absolutely nothing to do with how this technology is regulated.
    - Matticus_Rex 3y ago
      It's the most analogous process we have; the courts may not end up treating them the same, but it's unlikely they'll say it has nothing to do with how AI gets regulated.

nickelcitymario 3y ago

I may be wrong, but I believe they were just intending to summarize.

kemayo 3y ago

I'm quoting the relevant bit from Valve's email in the reddit post, to indicate exactly what they're actually forbidding. Notably: not all AI-generated content, but rather AI-generated content from models that were trained on material that's not owned by the person submitting the game.

XCSme 3y ago
Can that even be proven/tested?
- BlueTemplar 3y ago
  Depending on the court, burdens of proof can vary a lot...
  - dragonwriter 3y ago
    In court, it would be the entity claiming infringement that would have the burden of proof that their exclusive rights under copyright were violated by the assets in question, not the distributor of the asset that had the burden of showing ownership of every item in the training set of the model used in some part of the workflow.
    - kemayo 3y ago
      What I'd assume Valve is worried about is that it only takes one major decision against Stable Diffusion in court to suddenly leave us in a state where "this game used Stable Diffusion" is the proof that's needed. Given the whole "Stable Diffusion reproduces the Getty Images watermark" lawsuit[1] that's still ongoing, it's not an idle concern. [1]: https://www.theverge.com/2023/1/17/23558516/ai-art-copyright-stable-diffusion-getty-images-lawsuit https://www.theverge.com/2023/1/17/23558516/ai-art-copyright...
      - dragonwriter 3y ago
        > What I'd assume Valve is worried about is that it only takes one major decision against Stable Diffusion in court to suddenly leave us in a state where "this game used Stable Diffusion" is the proof that's needed. Hard to see how any plausible outcome that would have that result for users of SD (if model training isn’t fair use, that’s definitely a blanket-liability issue for Stability.AI — and Midjourney, and OpenAI, and lots of people training their own models, either from scratch or fine-tuning, using others' copyright-protected works. But “using a tool that violates copyright in the workflow” is not itself infringement; whether and in what situations prompting SD to produce output makes the output a violation of copyright (and whose) would be a completely different decision, and while Ibcan certainly see cases (such as deliberately seeking to reproduce a particulaflr copyright-protected element, like a character, from the source data) where it might be (irrespective of the copyright status of the model itself), I haven't seen anyone propose a rule that could be applied (much less an argument that would justify it as likely) based on copyright law that gets you to “used SD, in violation”. Lots of blanket ethical arguments about using it, but that’s a different domain than law.