If you read the article, you'd find they admit to that, their position is nuanced, and this is not a "gotcha".

> We'll be honest about where we stand. We don't like copyright as it exists. We'd like to see it radically reformed, and we wouldn't mourn it if it were abolished. Our community lives in its grey areas every day: many of the models on our sites are fan ports of characters from commercial games, made out of love and shared for free. When a studio sends us a takedown notice, we comply.

> While we as small artists and fan communities get copyright used against us, the world's largest companies are committing copyright infringement at a scale unseen in human history. This has been the case since the invention of the internet of course; digital media make duplication, remixing and resharing trivial, but copyright law never caught up. Filesharing stabbed copyright in the gut. AI is just twisting the knife, and completes the creation an unequal system where copyright infringement is only a crime if you're a low net worth individual.

Rich companies are infringing copyright at industrial scale in order to build their LLMs, and even if caught and challenged, they don't suffer any catastrophic punishment. The infringed party gets enough money to shut them up and everyone moves on. A fine is a price. They'll keep doing it with impunity. Meanwhile the small guy who doesn't have copyright lawyers on retainer is getting shafted, time and time again.

"When a studio sends us a takedown notice, we comply." Well, yes, obviously they have to, or risk not existing anymore.

Also, "fan ports" is doing a lot of the heavy lifting. They're assets extracted from commercial games, uploaded with whatever license the person who extracted and packaged them chooses.

From their argument, if they truly think copyright should be abolished, an LLM capable of remixing that work into whatever you want would be a good thing, right? Ultimately this would empower the end-users of the models.

There's an argument to be made that these models trained on the world's knowledge shouldn't be locked behind closed models controlled only by big corps, but in this case it's good that anyone can download data and train such a model, right?

I can't speak for roblox intentions, and I am well aware of their history. But so far they seem to have published at least some of their model work to HuggingFace: https://huggingface.co/Roblox/

What nuance am I missing?

> Rich companies are infringing copyright at industrial scale in order to build their LLMs, and even if caught and challenged, they don't suffer any catastrophic punishment

And the same thing is happening here? They’re hosting copyrighted materials and have also not suffered catastrophic punishment.

"It's OK if I do it, but not if someone more successful than me does it" is still blatant hypocrisy.