Twitter to label deepfakes and other deceptive media
1–10 of 151 posts
Re: Twitter to label deepfakes and other deceptive media
#2I'll keep an eye on this one. I'm hopeful because we need something like this, but man it's going to be hard with LOTS of problems on both sides of the line.
Guys who get labeled as fake who didn't deserve it and guys who should be labeled not getting it. All the false positive/negatives will have a terrible time.
Re: Twitter to label deepfakes and other deceptive media
#3This seems like an insanely hard problem with many false positives/negatives. But of course the media hysteria against tech and fake news might eat it up and praise twitter. Nice PR move for non techies.
Re: Twitter to label deepfakes and other deceptive media
#4This seems like an insanely hard problem with many false positives/negatives. But of course the media hysteria against tech and fake news might eat it up and praise twitter. Nice PR move for non techies.
Yeah I'm most concerned about the false sense of security this gives. It won't find deep fakes from unpublished methods.
Re: Twitter to label deepfakes and other deceptive media
#5Very interested to know how this would work technically, or if it is just another task for the army of human content moderators.
Is the end result of this just going to be that deepfake generators face natural selection pressure to improve beyond the exclusion threshold of this feature?
Re: Twitter to label deepfakes and other deceptive media
#6Would not surprise me when, eventually, browsers are going to mark content as fake. If there's DBs of ad networks, there might also be DBs of deepfakes.
Re: Twitter to label deepfakes and other deceptive media
#7This seems like an insanely hard problem with many false positives/negatives. But of course the media hysteria against tech and fake news might eat it up and praise twitter. Nice PR move for non techies.
> Nice PR move for non techies.
Depends if the praise they get for "at least they're doing something/not pretending the problem doesn't exist" outweighs the criticism they get for false positives, false negatives, inconsistency/bias, arbitrary judgements, response times, or labelling so many things fake that it becomes meaningless.
Way back in 1994, Time magazine published a photo of OJ Simpson that was photoshopped to make his skin darker [1] - but having some colour-correction of magazine cover photos is inevitable. The line between regular and deceptive colour correction is blurry, and some people will be upset no matter where you try to draw it.
[1] https://people.southwestern.edu/~bednarb/su_netWorks/project...
Re: Twitter to label deepfakes and other deceptive media
#8I have yet to see a deep fake that was passed of as real. Is there any major incidents of this happening that I'm not aware of?
Re: Twitter to label deepfakes and other deceptive media
#9I have yet to see a deep fake that was passed of as real. Is there any major incidents of this happening that I'm not aware of?
Re: Twitter to label deepfakes and other deceptive media
#10Would not surprise me when, eventually, browsers are going to mark content as fake. If there's DBs of ad networks, there might also be DBs of deepfakes.
How come this hasn't happened for news articles without video? Or has it? I am aware of the bias issue but that never stopped volunteers before.