Rather than training another AI network to detect fakes, I think we need to use semantic likelihood. That is, what is the probability of this picture being real based on its content? Low probabilities demand higher proof. I think this is what my brain does.
3 comments
[ 2.2 ms ] story [ 21.1 ms ] threadI will admit that I did have a puff but I did not inhale.
Anyway, my mates are pranksters.
Its turns out they bought a blow-up doll that looked exactly like the old trouble indoors.
It did look a bit fake to me and they were trying to convnce me that it was actually her.
So I stuck a pin in it.
It soon whent flat and it wasnt the old trouble
How to spot a deep fake?
Stick a pin in it