Posted by aliasxneo 1 hour ago
It would be interesting to know where the "corrected" deepfake originated.
This keeps happening because it works. But why does it work? Why isn't there blowback?
This has always been the case.
However, the more interesting thing going on is that we have never had to reckon with this problem anywhere near as much as we do now, since the barrier to creating fakes has been reduced from "skilled software user" to "all five billion humans who can write".
edit: the story was retracted so the "people manipulated with lies" are the people who believed oregon journalism project?
[0] https://openai.com/research/verify/
e: Okay, now I feel stupid for even bothering to do that. Look below her name and you'll see the leftovers of the old flag.
edit 2: Retracted. https://www.oregonjournalismproject.org/editors-note-october...
https://x.com/kgallella/status/2112332869461045733
https://x.com/Honest_Bender/status/2112301833708519659
Of course it doesn't conclusively prove that the flag was definitely right side up, but you'd really have to bend over backwards to explain why Oregon Journalism Project's "debunking" is watermarked but the counter-debunkings aren't.
And now that you’ve said that I’ve retried the same image 5 times and gotten AI 3 times and Not AI twice.
Something tells me that tools is about as reliable as Pangram.
https://www.google.com/imgres?imgurl=https://lookaside.fbsbx...
As an example, I can screenshot your comment and have AI edit it to have you say “I really enjoy misleading people on the internet for money”, post it, and then say “I took this screenshot before they deleted it.”
But also just look at the bottom-left of the image as is mentioned elsewhere in this comment thread: they didn't even bother to photoshop the entire image in their "debunk".