Hollywood actors and actresses lost their minds over the launch of AI actress Tilly Norwood. — Reuters
Well, the AI wars just got worse. Just when I thought the AI platformers had figured out how to temper their conquests and deliver tools that would result in long-term wins for everyone, OpenAI went and launched Sora 2, a one-stop shop for prompt-based short-video copyright infringement on the iPhone app store and it skyrocketed to number one like a bullet with 164,000 downloads in 48 hours.
If you were busy and missed the whole fiasco around OpenAI’s Sora private app release, you missed a parade of prompt-driven AI-generated short videos featuring Ronald McDonald fleeing the police in a burger-shaped car – along with all sorts of protected IP like Nintendo, Southpark, even the Simpsons characters doing whatever meme-able things the app’s invite-only users could unleash on an amused public.
Oh, the lawyers were not amused.
Especially those who work for the companies holding the IP copyrights for those mostly animated characters. Also unimpressed were the lawyers who work for famous people who were about to be placed in compromising positions – once some idiot decided the world would get a laugh out of seeing Taylor Swift dressed as a Nazi and waving a banana while shoplifting in an adult video store.
Yeah, the AI wars went nuclear. No sign on when they get better. Because maybe all OpenAI did was rip a page from the tried-and-true “apologise after instead of ask for permission first” playbook. But I believe they took that old chestnut of a strategy to the next level, using the Sora private release as a test of opt-out versus opt-in for AI.
Maybe they knew exactly what they were doing. And maybe they got us again. Reckless speculation to follow.
Modern generative AI is built on IP theft
Man, it sounds harsh when you say it out loud but there really isn’t a counter-argument anymore.
Back in 2010, I co-invented some of the first public-facing generative AI. However, unlike today’s version, our models were developed solely on the private data of our customers, data which never left their possession, giving them total control over how that data was exposed to either their own customers or the general public.
Now, if you decide that you want to sell AI not just to specific customers, but to the whole wide world, you’re gonna need – you guessed it – the whole wide world’s data.
How do you get all that data? Not only that, how do you get permission to use all that data?
Well, in my experience, you would need to scrape first and apologise later.
These shenanigans all came to a head in 2024, and by the end of the year, people like me were raising their hands and asking if we were just going to let everyone get away with the mass theft of all the world’s IP – while also noting that said IP was “housed” in a notoriously poor and unverified data store.
But the thing is, people like me already knew the answer, because while this was the first time we heard the opt-out/opt-in argument as it related to AI, it was the same opt-out/opt-in argument we had already heard about SEO. SEO was such a lure of cash to be made that we not only let the bots in, we tweaked the content and added the keywords to make it easier for the bots to collect whatever IP they wanted. Just give us that coveted high-ranking search link, please!
We let it be OK to do that.
Not only did the AI platformers ride the backs of that process, allegedly, but when they had what they needed, they broke the original promise. No more links for you!
What do we get in return? Apparently, users get a chance to “engage with their family and friends through their own imaginations.”
This is not about user imagination
It never was.
It’s the same story Facebook gave us when they started letting brands make social accounts, way back in the olden times.
It’ll be fun. You can interact with Diet Pepsi the same way you interact with your best friend from the third grade. Or your mom. It will deepen relationships with end users and the brands they love.
Come on.
It hasn’t stopped. Just recently, when Hollywood actors and actresses lost their minds over the launch of AI actress Tilly Norwood, the let’s-all-just-calm-down response from the creators was: “AI offers another way to imagine and build stories.”
I gotta call (expletive) here.
How is it not a way to generate content without paying the people who own the IP – in this case the actors and actresses Tilly was trained on?
Same thing here with Sora. This has, in my opinion, nothing to do with interacting with friends in a brand new way. It has everything to do with generating content without paying the creators of the IP.
Opt-out as policy is a joke
And it always has been.
Lawyers: Wait, you can’t do opt-out. It’s a completely onerous burden on the owners of the IP.
AI platforms: Oops. We just did. We’ll fix it. Here’s some money.
I think, at this point, generative AI might be serving as a testing ground for the next phase of AI – prescriptive, predictive, autonomous, and agentic, or the more “thinky” AI.
The rest of that “real” AI is where the “real” money is, but like generative AI, it doesn’t work without a lot of data, and in many cases, it doesn’t work well without proprietary data.
And paying the owners for that proprietary IP is really expensive.
I don’t think copyright law makes a dent here. Copyright law is nothing more than a warning and an excuse to start a legal battle. Besides, the toothpaste is already out of the tube. Because that’s not Spongebob Squarepants in that Sora video, it’s ShapeySlacks McPhilFish, or whatever derivative you want to slap on it.
But, that won’t stop the IP owners from trying. Last month, Rolling Stone-owned Penske sued Google. And in the same week Hollywood freaked out about Tilly Norwood, Disney sued Midjourney and sent a cease-and-desist to Character.
Welcome to the litigation phase of Generative AI, folks. If that’s where we’ve arrived then I don’t see it getting better any time soon. So protect your IP, because during the next phase of “real” AI, that proprietary data is going to be a lot more valuable. – Inc./Tribune News Service
