Beyond that, there are so many open questions. What is an expression? At what point do combinations of statistics no longer represent expressions? At what point does Copyright disappear? This paper doesn't address any of them, yet presents many of its conclusions as facts.
GENERATIVE AI
-
AI Training Fair Use Argument and Copyright Infringement Liability
By
–
On the U.S. side, the argument is a common one that's been made many times before:
"Training is Fair Use." Reproductions of protected works *are* made during training, so AI companies could be liable for infringement, but the argument goes, that's Fair Use. -
Copyright Exception 17 USC 117 Unlikely in AI Legal Battle
By
–
The basis for this is discussed in the technical section above. It's an argument that has also been made before already and contested, so there's not much new here either. 17 USC 117 is raised as a © exception, but frankly that's such a stretch there's no chance it'll fly.
-
Copyright paradox neural network weights inference time
By
–
The resulting paper, which was written since then, coincidentally advocates that there are no Copyrights on weights and that those are non-expressive representations, and yet those get reassembled magically into expressions that can be Copyrighted at inference time.
-
Legal battles over LLaMA weights and AI output ownership rights
By
–
VanL was hired by theshawwn to handle a DMCA claim by Meta about taking down the LLaMa weights, so they're trying to prove that weights are in the public domain. He was also hired by icreatelife to attempt to show that it's possible to own rights on the outputs of AI systems.
-
Compositing tools for AI-generated video content workflow
By
–
After effects or any other tool for compositing. I’m not sure you can do multi layer compositing in A1111 but @toyxyz3 might have another idea
-
Stochastic Parrots Paper Accessible via ACM Digital Library
By
–
Thank you, @cher0x801 Also — Stochastic Parrots was published as Open Access. There's absolutely no need to point to random storage locations. It will remain accessible at its point of publication in the ACM Digital Library:
-
Op-ed critique: ableism and AGI misconceptions in AI discourse
By
–
There's a lot I like in this op-ed, but unfortunately it ends with some gratuitous ableism (and also weird remarks about AGI as a "holy grail"). First, the good parts:
-
Segmentation technique pour optimiser ControlNet dans la génération d’images
By
–
One thing you can try is to just use segmentation to isolate the background and process it independent of the foreground — then composite them together. That way you can really refine the multi-controlnet settings for the background.
-
EbSynth temporal coherence and keyframe interpolation challenges
By
–
Agreed, temporal coherence leaves more to be desired. This is because the keyframes are not visually consistent, so when EbSynth is used to interpolate between them and crossfade you get jumps in the subject matter.