In what data centers would these open models be run such that copyright laws will not apply?
Serious question. Trying to figure all this out.
Is it that you think people will run the models on their own laptops or phones? Or will there be some offshore municipality where the models can be served from that is out of the reach of copyright laws? Do you have another idea in mind entirely? How are you thinking on all this?
However, the way the parent worded it ie $1 of compute electricity - requires the compute to exist and temporarily be loaned. Otherwise “just a $1” would require a massive capex to buy/build compute.
You could make them with public figures that aren’t alive any more for a day or two longer, but those get blocked now, too.
That said: I think you’re ultimately right, it’s just that it’s more of a past tense thing. My primary remaining fun use case is to push as close to the guardrails as possible to make embarrassing funny videos of my friends and family members, who are doing the same to me.
If you can’t generate videos of real people or movie characters getting arrested or doing insane things, no one is going to care about Sora anymore.
OpenAI knew that, played fast and loose with IP laws because… they wanted that bit of popularity to impress investors or something… then the lawyers got nervous and now they’re dialing it back down.
It’s a trick they can pull once but that’s it.
I suspect the more limited it inevitably becomes due to lawyers being lawyers, the more its popularity will wane.
Ironically enough, that’s why I think open-source models will still come out ahead in the long term. People really want to make videos with Pikachu in them.