excreationist
Married mouth-breather
I'd recommend this video from 9 months ago:
"like I said there have been experts that have been tripped up so many times recently with this kind of stuff"
So even the experts are having trouble... and 9 months ago is somewhat outdated now. Also some people are thinking that real videos were actually AI.
Spot all fakes is a very different challenge than spot any fakes.
On that show they had false positives and false negatives - maybe almost as good as chance.
Note I'm talking about the best generative AIs... some of them are nowhere as good. Apparently Sora 2 (before it was closed down) cost about $2 per 10 seconds (which I think OpenAI would normally pay). It's hard to offer that level of quality for free.I've caught many AI fakes because the AI has no real understanding of physics and the like (consider the scene, what forces created this? Where is the focal plane of the camera?).
I think it goes on whether people think they're fake without having any hint that it could be (like the host telling them that some are fake). If they are indistinguishable from reality I'd expect people on the show to guess half right and half wrong. It's related to a Bernie Sanders video I was watching on YouTube. I only suspected it was a fake because he said "a one zero year old boy" instead of "a ten year old boy". Even after that I couldn't see any signs that is was a fake but people on his subreddit page could tell.Sometimes they'll be right and you can't tell--but the AI isn't good enough to make them reliably correct.
On the other hand I guess being "indistinguishable from reality" suggests that the viewer is very suspicious though it could involve a person that is just taking it for granted that could notice significant issues if they arise.
Last edited: