Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

OpenAI Strengthens Sora Protections Following Celebrity Deepfake Concerns

I guess it's not entirely surprising that AI video generators like Sora at least initially allowed users to create videos containing the likeness of anyone, given that the whole LLM-based AI industry, for the most part, got its start by using copyrighted works and websites to train its models, without compensation to the people who created those works.

But once they were criticized for scraping everything they could, some AI companies started entering into licensing/compensation partnerships with various content creators/providers, and OpenAI now has more such partnerships, at 10-12, than the other AI companies combined. So you'd think OpenAI would have put these new restrictions on Sora from the start. But NO.
 
Strange how this became a popular app: when searching the AppStore the first results are:
AI Art: Photo Video Gen...
ChatGPT
VideoGPT - AI Video Ge...
Grok
Google Gemini


...and so on. How the hell am I supposed to install a popular app if I can't find it in the AppStore, Apple?
 
OpenAI knew what they were doing when they released this. It's insane to suggest otherwise.

They just steal, steal, steal to see what they can get away with. If they get their hands slapped hard enough, as always happens anyway, they put in place "guardrails". What b-****.

It's a shameful way to run a business with absolutely zero respect for people's intellectual property or livelihoods.
 
  • Like
Reactions: sickoagb and oldwatery
Well there better start getting at facebook then because their ai allows you to generate deep fakes of Celebrities via images. I have done a few (but never uploaded them to any site though)

Seems they need to be clapping down on these ai generator tools.
 
There of course is always going to be idiots that are going to use this to harm someone, that was never going to be in doubt, what it should have been done first and foremost is to put restrictions on the kind of use it can be used for in the first place. Adobe did this from the very beginning with their ai tools. Especially their video. So why didn't openai do this?

Meta needs to start doing this as well, the amount of F1 Brad Pitt images i generated alone should be concerning that is on my desktop background should be concerning :p (no he isnt naked)
 
OpenAI said there are "strong free speech interests" in depicting deceased historical and public figures, but authorized representatives or estate owners can request that their likeness not be used on Sora cameos.

Honestly, it should be the other way around. There's no "free speech interest" in putting people in fake AI videos, whether they're dead or alive.
 
  • Like
Reactions: oldwatery
Privacy is the most important requirement in apps like these. Without proper control, more such deepfakes will be generated.
 
  • Like
Reactions: mganu
goobot said:
In all honesty ai like this shouldn’t be made available to the public, the only purpose of this app is to make fake videos of people, like what positive affect can that have on society.
Click to expand...
I think vice versa: we need to make that stuff everywhere so people will get tired from trying to fake reality and will instead pursue real things.

This is happening in photography btw. For the last 10 years people have been chasing unrealistic colors, sharpening, noise reduction (why remove noise? Those are photons, literal physics in front of you! And every camera has their own noise pattern), HDR. And now we see more gen Z people buying or using old digital cameras or even going film, because of more realism
 
goobot said:
In all honesty ai like this shouldn’t be made available to the public, the only purpose of this app is to make fake videos of people, like what positive affect can that have on society.
Click to expand...
I disagree with the AI-fakes panic.
Fake videos and pictures are already easily available, reasonable people already know all sources must be checked.
While the others already easily fall for "FWD: FWD: FWD: FWD: haitians eat kitten-burgers". These people are the real target for misinformation.
The greatest global political shift since WW2 has already happened because of this.
 
Why do people care so much about deepfake pornography, if we all know that it is not real? I would find it funny, if such a video of me was created. In the video I would probably look much better than in reality.
 
And what if you just happen to be born looking exactly like Taylor swift or someone else famous? Are you barred from making your own?
 
The core problem here is that these are just public facing safeguards. The capacity still exists and OpenAI will either cozy up with bad faith actors directly or the bad faith actors will catch up to OpenAI's capacity. News stories like this feel a lot like bailing water out of the Titanic after it struck the iceberg. The fact that whole cloth simulation of real people is becoming more believable to do with minimal investment by the prompter has huge societal concerns and a few validations in an API isn't going to save us from the impact.
 
  • Like
Reactions: Moi Ici
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back