2 ms·

Deepfake videos are getting good

4 points by bentocorp 2y ago

ChrisArchitect 2y ago
Earlier: https://news.ycombinator.com/item?id=42930639 https://news.ycombinator.com/item?id=42930639
mojomark 2y ago
I don't understand why prominant people, who are in the public eye (politicians, pop artists, other narcissists) don't just digitally sign their content and maintain a crosse-referenceable ledger (on a site they or trusted 3rd party control) of specific locations where they've authorized their content to be posted, so that authorship can be verified (unless the cross-reference hosting site gets hacked, but let's assume that's very secure). I mean, this seems like a no-brainer, so am I missing a fundamental point regarding public key encryption? (very possible - I'm not a crypto person, I'm legitimately asking because I've wondered about this for a long, long, long time) It's the digital version of saying to someone: "hey ding dong - did you authorize this craziness?" Then them saying: "no way, of course not, that shit's whack"
- yorwba 2y ago
  The deepfakes that people are worried about are ones that, if they were real, they'd not authorize anyway, so the lack of a digital signature wouldn't help distinguish deepfakes from leaks. At the same time, digitally signing something real just makes it hard to disavow later when you suddenly feel embarrassed by it. Hence there's little reason to sign anything. On the other hand, maintaining a list of photos that are allowed to be posted, where platforms take down anything else with a facial recognition match might work better than the status quo where platforms maintain lists of stuff that isn't allowed to be posted. But it's up to platforms whether they want to implement such a system.
  - mojomark 2y ago
    OK, so, extend that philosophy to any media content. The method I describe above would force whomever is publishing content that that want to "seem legitimate" assign a human person (digital signature) as the source. Then, it would be easy to say "well, this content doesn't have a legitimate source that I can point to, so, it's probably fake. And, if people are afraid to post the content without personally backing it up, they probably shouldn't be posting.

4 points by bentocorp 2y ago

Earlier: https://news.ycombinator.com/item?id=42930639 https://news.ycombinator.com/item?id=42930639

I don't understand why prominant people, who are in the public eye (politicians, pop artists, other narcissists) don't just digitally sign their content and maintain a crosse-referenceable ledger (on a site they or trusted 3rd party control) of specific locations where they've authorized their content to be posted, so that authorship can be verified (unless the cross-reference hosting site gets hacked, but let's assume that's very secure). I mean, this seems like a no-brainer, so am I missing a fundamental point regarding public key encryption? (very possible - I'm not a crypto person, I'm legitimately asking because I've wondered about this for a long, long, long time) It's the digital version of saying to someone: "hey ding dong - did you authorize this craziness?" Then them saying: "no way, of course not, that shit's whack"

yorwba 2y ago
The deepfakes that people are worried about are ones that, if they were real, they'd not authorize anyway, so the lack of a digital signature wouldn't help distinguish deepfakes from leaks. At the same time, digitally signing something real just makes it hard to disavow later when you suddenly feel embarrassed by it. Hence there's little reason to sign anything. On the other hand, maintaining a list of photos that are allowed to be posted, where platforms take down anything else with a facial recognition match might work better than the status quo where platforms maintain lists of stuff that isn't allowed to be posted. But it's up to platforms whether they want to implement such a system.
- mojomark 2y ago
  OK, so, extend that philosophy to any media content. The method I describe above would force whomever is publishing content that that want to "seem legitimate" assign a human person (digital signature) as the source. Then, it would be easy to say "well, this content doesn't have a legitimate source that I can point to, so, it's probably fake. And, if people are afraid to post the content without personally backing it up, they probably shouldn't be posting.