It’s telling to how society values copyright of different media that 4 years into people yelling about these being copyright violation machines the first time there’s been an emergency copyright update has been with video.
So people who spend time working on code or art should have exactly zero protection against somebody else just taking their work and using it to make money?
No, but the current system is totally idiotic. Why not have a fixed timeframe i.e. 30-50 years to make money? Life of the author + x years is stupid not only because it's way too long, it keeps going until way after the creator is no longer benefitting, and it can cause issues with works where you don't know who the author is so you can't cleanly say it's old enough not to have copyright.
I'm not sure for most (specifically smaller, who need the most protection) creators this would actually change very much. Media typically makes money in it's first few years of life, not 70 years on.
If you change it to 30-50 years then nothing changes for Sora.
Most of the (obviously infringing) video I've seen is stuff well within the past 30-50 years.
Also, no copyright doesn't mean information free for all.
I still can't generate a fake video, pretend it's real, and then claim you're a criminal or you eat babies or something. Because that's libel.
Why do we have libel laws? Because the alternative is that you piss off Walmart or something and then they go and tell all employers that you're a pedophile and then you starve to death, with the cherry on top of having a tainted legacy. So we definitely need the libel laws.
But that's a problem. The whole premise of Sora is that it generates fake video with the intention of making it as real as possible.
No matter how you cut it, Sora's business model, and moral model, is on shakey grounds.
This does not demonstrate a sound understanding of how the public domain works, why copyright lengths have been extended so ferociously over the last century (it's shareholders who want this), nor the impact it has both on creative process and public conversation.
This is a highly complex question about how legal systems, companies, and individual creatives come in conflict, and cannot be summarized as a positive creative constraint / means to celebrate their works.
- owners of large platforms who don't care what "content"[0] is successful or if creators get rewarded, as long as there is content to show between ads
- large corporations who can afford to protect their content with DRM
Is that correct?
Do you expect it to play out differently? Game it out in your head.
The vast majority of DRM is cracked very quickly; the only reason DRM cracking tools aren't more widespread is because of copyright law and the idiotic anti-circumvention provisions.
Consider that even DRM'd content is on torrent sites within hours of release.
"Hi, as the company that bragged about how we had ripped off Studio Ghibli, and encouraged you to make as many still frames as possible, we would now like to say that you are making too many fake Disney films and we want you to stop."
These attempted limitations tend to be very brittle when the material isn’t excised from the training data, even more so when it’s visual rather than just text. It becomes very much like that board game Taboo where the goal is to get people to guess a word without saying a few other highly related words or synonyms.
For example, I had no problem getting the desired results when I promoted Sora for “A street level view of that magical castle in a Florida amusement area, crowds of people walking and a monorail going by on tracks overhead.”
Hint: it wasn’t Universal Studios, and unless you know the place by blind sight you’d think it had been the mouse’s own place.
On pure image generation, I forget which model, one derived from stable diffusion though, there was clearly a trained unweighting of Mickey Mouse such that you couldn’t get him to appear by name, but go at it a little sideways? Even just “Minnie Mouse and her partner”? Poof- guardrails down. If you have a solid intuition of the term “dog whistling” and how it’s done, it all becomes trivial.
Absolutely. Though the smarter these things get, and the more layers of additional LLMs on top playing copyright police that there are, I do expect it to get more challenging.
My comment was intended more to point out that copyright cartels are a competitive liability for AI corps based in "the west". Groups who can train models on all available culture without limitation will produce more capable models with less friction for generating content that people want.
People have strong opinions about whether or not this is morally defensible. I'm not commenting on that either way. Just pointing out the reality of it.
It's a matter of time. I imagine they'll get more effect suppressing activations of specific concepts within the LLM, possibly in real time. I.e. instead of filtering prompt for "Mickie Mouse" analogies, or unlearning the concept, or even checking the output before passing it to user, they could monitor the network for specific activation patterns and clamp them during inference.
I can get it to do rides at Disney World (including explicitly by name) but it’s incredibly good at blocking superheroes. And that’s gotta be a pretty common prompt, yet I haven’t seen that kind of content in the feed, either.
And not just by name. Try to get it to generate the Hulk, even with roundabout descriptions. You can get past the initial (prompt-level) blocking, but it’ll generate the video and then say the guardrails caught it.