Photo-realistic lip-sync from text
ritheshkumar.com
ritheshkumar.com
I don't think that's a good thing, though. Right now you can catch a politician in a lie, on tape. Once that's easily faked people will just see whatever confirms their preexisting bias.
It's something more than being a skeptic - when there's no definitive proof of anything, you just pick and choose what you believe.
People who want to believe "fake news" will believe it, people who don't care, won't care, and people who think it sounds unlikely or don't want to believe it will either disbelieve it or hopefully attempt to fact check it.
To take a more positive view of this kind of tech, I think it has promise in producing satirical content.
??? Documentary = non-fiction, isn't it? I mean, would you use CGI to show someone doing something you didn't quite record them doing?
Fixed that for you.
https://www.thedailybeast.com/nasa-denies-that-its-running-a...
It was even picked up by space.com
https://www.space.com/37366-mars-slave-colony-alex-jones.htm...
Because it looks more like, and is seemingly marketed as - if a video of Barack Obama saying things he hasn't said is typical - a way to make prank videos.
Edit: And why is it even called ObamaNet? Is it endorsed by the former president?
(I'm not making a political point here, I'd ask the same if it was TrumpNet or GagaNet. Is Obama notably connected with this kind of tech or research?)
Very interesting, thanks.
If he had access to this, he could have released video reviews, which could have drawn a much larger audience.
Not a fan of the ObamaNet naming either - really seems to miss the mark.
Society as a whole will just have to realize that video is as alterable as text.
[0] https://www.nytimes.com/2017/06/09/insider/a-photo-of-james-...
From 1997, 20 years ago: https://www.youtube.com/watch?v=q-ix5IYz0cc
People are gullible enough as it is (e.g. 9/11 conspiracists), and usually just want to believe what they already think is true. This will just fuel ignorance and is another reason why it's very important that we somehow as a society get people to think for themselves sensibly.
That said, the demo they showed is very impressive, very good work by them.
Would the best approach to actually deceiving the viewer be a voice impressionist paired with this technique?
This is a Photoshop of voices and as such it is neither more or less dangerous.
We desperately need better ways to detect BS built into browsers, social networks etc.
Lyrebird.ai similarly followed this approach as they realized their own technology could be used for nefarious purposes given the current political climate.
I hope every company that's working on this stuff either fails, financially or technologically, or ideally gets regulated out of existence, at least until they commit to working on technology, policy and journalistic techniques to mitigate the absolutely inevitable misuse, because as a society we simply are not ready for this. If you work for one of these groups, you should either ensure they're working on said mitigations, or quit.