How did those releasing the dataset engage in a good faith debate? Is it possible for such a debate to happen after the fact?
@alexjc
-
Copyright concerns and DSM compliance in AI development
By
–
You're willingly trying to destroy Copyright, it's fair to expect an adversarial response. The DSM exists, your interpretation of it is not grounded in reality.
-
Fair Use Rules in AI: Balanced Perspective on Rhetoric
By
–
What rethoric in particular? I have tried to refine my views to be some of the most balanced and reasonable of almost anyone in this debate. Fair Use serves a purpose and is fine to rely on if you follows its rules, and abide by those expectations.
-
Fair Use Criteria and AI Research Innovation Beyond Scaling
By
–
I wouldn't call it theft because it's inaccurate. Transformative is only one of the criteria that must be fulfilled in order to obtain Fair Use. Sad to see you think brute force scaling is the only way forward for research; shows a lack of imagination and innovation.
-
Artist Copyright Infringement in AI Training Datasets
By
–
The artist's works were in the dataset, it's a Reproduction right needed during training. Thus, infringement.
-
Infringement Methods and Uninformed Online Behavior Patterns
By
–
There are as many ways to infringe as there are ways to be the uninformed Reply Guy.
-
CEO’s No-Copyright Content Feature Raises Ethical Concerns
By
–
Your CEO advertised no-Copyrighted content as a feature… (IMHO that's incorrect.) The logical implication is that both of you think it's OK, otherwise your company wouldn't commercialize it! Understood about no debates, here's my final statement:
-
Fine-tuning Models Cannot Bypass Copyright Infringement Claims
By
–
If your startup reproduces a "Belgian School Comics" fine-tune hoping get around a lawsuit for #Samivel's Copyrighted content, it's not going to work. That content is in the original base model and you're *intentionally* trying to replicate it (+PR) so it's still infringement…
-
Stable Diffusion Fine-tuning Copyright Dataset Concerns
By
–
You're fine-tuning SD, right? You're aware that Samivel is in the original datasets too, yes? Why are you trying to avoid Copyrighted images on the fine-tune only? Might as well own it…