5 ms·

Has there been a law suit filed on the image generation side? Dalle and Stable Diffusion trained on images on the web, a lot of which weren't even freely licens

Has there been a law suit filed on the image generation side? Dalle and Stable Diffusion trained on images on the web, a lot of which weren't even freely licensed. So I would think similar legal arguments would apply there.

Karunamon 4y ago
The big difference there is that you never get a 1 to 1 copy of the source content out of the image models, where you often do with copilot. Whether your use is transformative is a part of the fair use legal test.
- rvz 4y ago
  Except that not only Stability AI have already admitted to training on copyrighted images without the permission or attribution of many artists [0], they already have a commercial SaaS API platform which uses the model [1]; effectively throwing out the sloppy 'fair use' or 'transformative' purposes claims as any artist can see all the digital art that Stability trained on without permission or attribution [2] and even producing outputs of famous copyrighted images verbatim. There is no difference between the two. Given that another project called Dance Diffusion was trained on public domain music and audio and permission by musicians, it is clear that Stability knew that they would be sued to the ground if Dance Diffusion was trained on copyrighted music. [3] They already seem to have admitted and felt guilty into trampling over the copyright and watermarks of images and knowingly avoided doing the same with music. My point is, use and train on public domain content only and content that has the permission of the creator. This applies to all of them; DALL-E, Copilot, Stable Diffusion. Clearly wasn't a problem to use public domain music with Dance Diffusion was it? [0] https://venturebeat.com/ai/stability-ai-to-honor-artist-opt-out-requests-for-stable-diffusion-3/ https://venturebeat.com/ai/stability-ai-to-honor-artist-opt-... [1] https://platform.stability.ai/ https://platform.stability.ai/ [2] https://twitter.com/EMostaque/status/1603147709229170695 https://twitter.com/EMostaque/status/1603147709229170695 [3] https://techcrunch.com/2022/10/07/ai-music-generator-dance-diffusion/ https://techcrunch.com/2022/10/07/ai-music-generator-dance-d...
- Kiro 4y ago
  > where you often do with copilot Not at all. You need to bait it really hard and push it into a corner for it to reproduce anything. At that point you might just as well go to the repo and copypaste the code directly. I've used Copilot since day one and still haven't seen anything that felt like a 1 to 1 copy of something. It's highly contextual and all about using the code I have already written to craft its suggestions. Even if I ask it explicitly for a known algorithm it will use my code style, patterns and naming conventions to write it.
- ROTMetro 4y ago
  I seem quite a few exact copies of watermarks transormatively 'artistically generated' on AI images.
  - Karunamon 4y ago
    An identical/similar blob of pixels across tens of thousands of otherwise unrelated images is exactly the kind of thing I would expect to see mashed up. I'm not sure what your argument is here.
sfifs 4y ago
Definitely about to happen as soon as lawyers can establish reasonable damages can be assessable to be worth the effort - or they piss off some rich artist who will be willing to underwrite the cost ala Theil did with Gawker. There was a ruckus just some weeks ago about some artist feeling that her style of art has become very easy to rip off.

Karunamon 4y ago

The big difference there is that you never get a 1 to 1 copy of the source content out of the image models, where you often do with copilot. Whether your use is transformative is a part of the fair use legal test.

rvz 4y ago
Except that not only Stability AI have already admitted to training on copyrighted images without the permission or attribution of many artists [0], they already have a commercial SaaS API platform which uses the model [1]; effectively throwing out the sloppy 'fair use' or 'transformative' purposes claims as any artist can see all the digital art that Stability trained on without permission or attribution [2] and even producing outputs of famous copyrighted images verbatim. There is no difference between the two. Given that another project called Dance Diffusion was trained on public domain music and audio and permission by musicians, it is clear that Stability knew that they would be sued to the ground if Dance Diffusion was trained on copyrighted music. [3] They already seem to have admitted and felt guilty into trampling over the copyright and watermarks of images and knowingly avoided doing the same with music. My point is, use and train on public domain content only and content that has the permission of the creator. This applies to all of them; DALL-E, Copilot, Stable Diffusion. Clearly wasn't a problem to use public domain music with Dance Diffusion was it? [0] https://venturebeat.com/ai/stability-ai-to-honor-artist-opt-out-requests-for-stable-diffusion-3/ https://venturebeat.com/ai/stability-ai-to-honor-artist-opt-... [1] https://platform.stability.ai/ https://platform.stability.ai/ [2] https://twitter.com/EMostaque/status/1603147709229170695 https://twitter.com/EMostaque/status/1603147709229170695 [3] https://techcrunch.com/2022/10/07/ai-music-generator-dance-diffusion/ https://techcrunch.com/2022/10/07/ai-music-generator-dance-d...
Kiro 4y ago
> where you often do with copilot Not at all. You need to bait it really hard and push it into a corner for it to reproduce anything. At that point you might just as well go to the repo and copypaste the code directly. I've used Copilot since day one and still haven't seen anything that felt like a 1 to 1 copy of something. It's highly contextual and all about using the code I have already written to craft its suggestions. Even if I ask it explicitly for a known algorithm it will use my code style, patterns and naming conventions to write it.
ROTMetro 4y ago
I seem quite a few exact copies of watermarks transormatively 'artistically generated' on AI images.
- Karunamon 4y ago
  An identical/similar blob of pixels across tens of thousands of otherwise unrelated images is exactly the kind of thing I would expect to see mashed up. I'm not sure what your argument is here.

sfifs 4y ago

Definitely about to happen as soon as lawyers can establish reasonable damages can be assessable to be worth the effort - or they piss off some rich artist who will be willing to underwrite the cost ala Theil did with Gawker. There was a ruckus just some weeks ago about some artist feeling that her style of art has become very easy to rip off.