OpenAI’s Sora 2 lets users insert themselves into AI videos with sound

Surely this will be the AI product, feature, or service to become profitable and pay back the hundreds of billions that have been invested so far.
Why would they need that ? They just have to keep investors convinced it's a good idea. As long as people imagine LLMs have value ... they will have.
 
Upvote
8 (8 / 0)

Fred Duck

Ars Tribunus Angusticlavius
7,437
Benj Edwards said:
In particular, OpenAI has built in layers of security for the cameos feature. It says that users can maintain control over their uploaded likeness: They can decide who can use their cameo in videos and can revoke access or remove videos containing their likeness at any time. Users can also view all videos containing their cameo, including drafts created by other people.
Wait, what? When you upload a cameo of yourself, you can set permissions to allow anyone to use it? In what universe is that a good idea? Shouldn't it simply be you can grant permission to individuals?

If I create a cameo of someone standing in front of a flat green backdrop saying something controversial ("Liquorice is delicious!") and screen record that, them revoking privileges is akin to closing one barn door after all the animals have fled.
 
Upvote
15 (15 / 0)
It's kind of sad that I completely lost the ability to believe any given picture is real on the internet unless there are other angles of it. There's always this thought in the back of my mind "what if it's some generative garbage?". That's awful.

If there's a way for the rich to get richer from it, they pollute anything and everything. They've already done it to the air we breath, the water we drink, put microplastics and forever chemicals in our foods and our own bodies, and got the ball rolling on the Kessler syndrome. Of course polluting our sense of reality was next.
 
Upvote
31 (35 / -4)

Fatesrider

Ars Legatus Legionis
25,499
Subscriptor
It's kind of sad that I completely lost the ability to believe any given picture is real on the internet unless there are other angles of it. There's always this thought in the back of my mind "what if it's some generative garbage?". That's awful.
You know, this COULD become a good thing, though totally the wrong way to do it.

In combating misinformation of any kind, especially biased news sites and such, disbelief is mandatory. You CAN'T TRUST WHAT YOU READ. But people do.

Now, you can't trust what you SEE, and people are beginning to realize that.

If we can get "people" to FINALLY connect that they can't trust what they read (at least without followup and some basic fact checking) with not trusting what they SEE, it COULD potentially begin to combat misinformation on a larger scale.

Granted, catering to biases will always be a thing, and both methods of misinformation do that. But with the tools to put the reader themselves into the picture, and then showing them that, and explaining how easy it is to fake something with images AND words, then it could raise awareness of how easy it is to take advantage of people's credulity to make them think and do things that are very much against their own self interests.

I'm NOT holding out much hope for that, because people are fucking stupid overall. Still it could help raise enough awareness for those who have SOME semblance of a clue and start to undermine the propaganda machines to a larger and larger degree.

After all, in a war, both sides use the most powerful weapons they have to achieve their goals. So you can't say "this side does it and that side doesn't do it" for very long before that position becomes entirely undermined and lost. The best you can hope for is pointing to both sides and saying, "don't pick a side that uses bullshit to convince you they're right, or that the other side is wrong."

Then let the cards fall where they may. I don't see living in a liberal utopia being viable, nor living in a conservative hell, either. Push people toward the center, where the leftist and rightist biases are least likely to be expressed, and you'd get a much more stable and peaceful society, I'd think.

it may not be heaven, but it wouldn't be hell, either.
 
Upvote
-15 (5 / -20)

Edzo

Ars Praefectus
4,443
Subscriptor++
The social media aspect of it is a little weird. Is there a big audience of people who want a TikTok style app made up exclusively of AI-generated videos? From yesterday's WSJ article, it mentioned it's supposed to encourage adult users that have been scrolling too long to make a video. I suppose that's one way to get people to close the app.
 
Upvote
18 (18 / 0)

forkspoon

Ars Scholae Palatinae
1,085
Subscriptor++
It's kind of sad that I completely lost the ability to believe any given picture is real on the internet unless there are other angles of it. There's always this thought in the back of my mind "what if it's some generative garbage?". That's awful.

It’s the post-evidence future we’ve all been waiting for! (and dreading)
 
Upvote
19 (19 / 0)

Tochoa

Smack-Fu Master, in training
64
Moderately impressive video at first glance, though it breaks down when you're looking at details like all of this stuff. The crazy amount of jump cuts in the intro video is definitely an intentional choice to make it harder to identify glitchiness. It's hard to tell if Altman is superimposed on the video or also generated. Probably could tell for sure if they had used a human instead.
 
Upvote
7 (8 / -1)
OMG! The rowing is out of sync, the beads of rain/water on the "blade" are perfection of spacing, the hairlines on the Vikings are clearly ... duplicates. Even the missing artwork on the shield.
However the moves of the bar gymnast are really good but the camera tracking is not good...it jerks at two movements indicative of a computer-generated camera track. Human cameraman doesn't move like that, nor those "action" remote cams.
 
Upvote
5 (6 / -1)

josephhansen

Ars Centurion
361
Subscriptor
OMG! The rowing is out of sync, the beads of rain/water on the "blade" are perfection of spacing, the hairlines on the Vikings are clearly ... duplicates. Even the missing artwork on the shield.
However the moves of the bar gymnast are really good but the camera tracking is not good...it jerks at two movements indicative of a computer-generated camera track. Human cameraman doesn't move like that, nor those "action" remote cams.
The gymnast video was a threshold for me- it's the first time I've looked at something AI generated, closely, multiple times, and not been able to say "yep, that's AI". The other videos, yeah, you can tell, but this one honestly looks like a real recording with mediocre camera work to me. I don't have words for how this makes feel, but I can try: "not good"
 
Upvote
8 (9 / -1)

Missing Minute

Wise, Aged Ars Veteran
1,386
The gymnast video was a threshold for me- it's the first time I've looked at something AI generated, closely, multiple times, and not been able to say "yep, that's AI". The other videos, yeah, you can tell, but this one honestly looks like a real recording with mediocre camera work to me. I don't have words for how this makes feel, but I can try: "not good"
That's because there's lots of fast motion, scrub through it and you'll find some blatant flaws.
1759350874408.png

If you were to watch in slow motion the bottom one would be particularly telling, the landing is missed and yet the routine continues.
 
Upvote
18 (19 / -1)

Kenjitsuka

Ars Scholae Palatinae
1,302
avoid common social media pitfalls like doomscrolling and addiction with what it calls a "new class of recommender algorithms"
Another boldfaced lie! If you want to prevent those things you DON'T include ANY recommendation algorithms! Let users search for stuff if they want to see it, it's just that easy!!!
 
Upvote
10 (13 / -3)

Brendan McKinley

Ars Praetorian
475
Subscriptor++
OpenAI claims it has designed the new app to avoid common social media pitfalls like doomscrolling and addiction with what it calls a "new class of recommender algorithms" that users can control through natural language instructions, rather than relying on traditional engagement metrics.
For some reason the above claim makes me think of a quote from Ray Anderson being interviewed in the 2003 documentary The Corporation which I've paraphrased below. Can't think of any reason it just popped in my head, none at all.

"...Can you make landmines sustainably? Well I don't think so. There's a more fundamental question that some products will not be made at all in a sustainable world."
 
Upvote
2 (2 / 0)

0101010111

Seniorius Lurkius
45
with mediocre camera work to me. I don't have words for how this makes feel, but I can try: "not good"
This is going to be abused by repressive governments to dismantle whatever’s left of democracy through disinformation and fakes, so that legit people spend their time defending themselves that they didn’t do whatever will have been doctored in the videos.
The Killer app
 
Upvote
0 (0 / 0)

FranzJoseph

Ars Centurion
2,895
Subscriptor
I can only commend the techbros for finally solving the Fermi paradox in our lifetimes. What an incredible achievement! Just too bad we won't be there to enjoy it, but I guess a few organ transplants will surely grant the select few the longevity, perhaps even "immortality", to at least be the last ones to shut down the lights?
 
Last edited:
Upvote
2 (2 / 0)

Don Reba

Ars Praefectus
3,349
Subscriptor++
That's because there's lots of fast motion, scrub through it and you'll find some blatant flaws.
View attachment 119380
If you were to watch in slow motion the bottom one would be particularly telling, the landing is missed and yet the routine continues.
It's amazing, how much we miss in motion.
1759360899177.png
 
Upvote
20 (20 / 0)
I love how every AI product is capable of making content on the spectrum of "mildly amusing in a distracting sort of way" to "incredibly dangerous and damaging"

It's hard not to be negative when it's all just so heavily biased towards negative outcomes. Like this video model has the possibility, at best, to make a video that makes someone go, "heh", and then they scroll past. And at worst be used to mass-manufacture propaganda.

Like yeah someone used it to frame their political enemies for a crime in the court of public opinion but on the up side it briefly amused a bored teenager so we can't stand in the way of progress!

The other day I saw Reese Witherspoon (who got really into NFTs, fyi) saying that ai is empowering for female filmmakers. Ok fine, I disagree, but let's assume that's true for the sake of argument. Does that theoretical empowerment even mildly offset the awful uses of ai against women? The fact that it's trivial to generate and share explicit material of anyone you have a picture of? Or that the ability to generate 'perfect' characters will have negative impacts on female actors? Or that it will further continue to warp beauty standards beyond their already absurd state?

Do even the theoretical positives outweigh the already demonstrably negative outcomes???
 
Upvote
17 (18 / -1)
The gymnast video was a threshold for me- it's the first time I've looked at something AI generated, closely, multiple times, and not been able to say "yep, that's AI". The other videos, yeah, you can tell, but this one honestly looks like a real recording with mediocre camera work to me. I don't have words for how this makes feel, but I can try: "not good"
At the end of the “routine”, the gymnast does a dismount on to the beam itself and then just sort of stands there.
 
Upvote
5 (5 / 0)

e.halap

Wise, Aged Ars Veteran
187
They could have used a GIF with a poo and some curly lines waving to suggest steam to represent Altman in the promo video. I heard they're a bit strapped for cash, this could have saved 40 bucks on processing power and they could reach 100 trillion dollars more quickly.
And maybe then he'd shut the fuck up for a while.
 
Upvote
-3 (0 / -3)