edit: I’ll take your knee-jerk DV, and any others, as an admission of an inability to speak to positive utility of this technology.
edit: I’ll take your knee-jerk DV, and any others, as an admission of an inability to speak to positive utility of this technology.
1. CGI for video editing - lower the bar of entry to de-age actors, or use a stand-in. Actor can't make it to a shoot that day? No worries, replace their face in post easily.
2. Identity protection - Cold call with someone that reached out to you, you're not sure if they're safe or dangerous, could be a good way to protect yourself.
3. Social media content for clients - become a fake avatar for hire essentially, customize your narrator for any video or brand. Video call centers with fake video (they already have voice modifiers and fake names), Enhanced VTuber sort of things (virtual avatars for streaming).
4. Unexpected outcomes: for example Holly Herndon created (and sold) access to an AI replica of her singing voice (n1), and I could see artists selling or renting access to their faces.
Obviously this can and will be used maliciously, but I personally could see myself using it for more positive reasons.
n1. https://holly.mirror.xyz/54ds2IiOnvthjGFkokFCoaI4EabytH9xjAY...
With that said almost every use-case cited was financial or monetary gain whereas I enquired about social utility and value.
That dishonesty ie creation of a fake avatar is cited as being of social utility strikes me as a reach. I don’t see how adding more dishonesty and facades to the world adds social value, but then I may just be of limited imagination.
Oh, right. I think the unfortunate answer is that nobody cares about that distinction any more and only about the former.
It could allow people to petition their government [0] without having someone bust down their door at 3am and disappear forever.
[0] Which is apparently important enough to have been included in the bill of rights
Surely just wearing a mask - virtual or otherwise - does the job without the deep fakery?
And having subtle clues from watching a person’s face while they’re proposing whatever action to overthrow the patriarchy is more convincing than some random person wearing a Guy Fawkes mask talking about revolution for the umpteenth time.
- Representing assistive robots/software with friendly human faces
- Reconstructing the likeness of people with permanent facial injuries when connecting with family
Other, questionably "legitimate", commercial uses are already in production:
- auto-generated corporate training videos
- "Personalized" advertising
I'm hating it already.
I could see it being used in AR to conceal identity to facilitate more equitable medical outcomes, I suppose.
Thank you again for the input! I was honestly at a loss for positive applications outside of financial gain.
I haven’t seen any ads driven by deepfake, or at least I don’t think I have. That advertising bit does sound rather obnoxious though!
"Deepfake therapy" lets therapists simulate the presence of dead or non-cooperative people [1]. A study showed positive results when sexual violence victims could safely discuss with a deepfaked version of their abuser [2].
[2]: Initial development of perpetrator confrontation using deepfake technology in victims with sexual violence-related PTSD and moral injury: https://www.frontiersin.org/articles/10.3389/fpsyt.2022.8829...
[0] https://collider.com/trey-parker-matt-stone-almost-made-sass...
I suspect it's going to become popular for both consensual-deepfake of oneself (PR, magazines, actors, pop stars, any form of public speaker) and "bought out" deepfake (actors selling out their image rights and then losing creative control; dead actors, etc.)
The political-deepfake is really going to accelerate the debate over how much free speech permits you to just lie about people, though.
Another analogy. Say somebody makes some hacking kit. Say it uses zero day exploits to compromise Windows, Mac, and Linux. Would any of us take issue with that? Would it be a different story if it was made into a push-button tool like WinNuke was in the 1990s? Or automated to the extent that somebody who can make a word doc could employ it against your systems? Is there really no feasible line of distinction here, in your eyes?
Think about it: people choose to trust or not trust based on a face. When deepfaking becomes a tool easily available to every average joe, appearance will lose some of its power. People will learn to lose their irrational trust in face.