3 ms·
The only mistakes I made were calling real ones fake, but I was always able to detect the fakes. Some of the audio sounded too clean but turned out to be real.
by jedberg 5y ago
The only mistakes I made were calling real ones fake, but I was always able to detect the fakes. Some of the audio sounded too clean but turned out to be real.
I guess I'm just extra skeptical.
- jacquesm 5y agoThat's an easy enough to achieve result though: simply label everything fake.
- jedberg 5y agoI mean sure, but I was actually trying. :P
- jacquesm 5y agoI was trying as well... 1 wrong (the penultimate one), and a perfect score for the remainder. I found the 'silent' clips the hardest.
- Shared404 5y agoI did much worse than that, but also found the silent clips the hardest. I did much better on them once I realized that the real speakers were much more likely to look around the room, or especially down toward the podium.
- jacquesm 5y agoA much easier criterium: the real speakers were speaking continuous sentences with logical spots for breathing pauses, the fakes discrete words. And I guess that's also why I had a harder time with the silent clips, then you don't have that signal.
- godelski 5y agoProblem is that you're primed to look for fakes. The game is much harder when your in a normal environment.
- brewdad 5y agoIt would be interesting to test how well any of us could spot these fakes if we weren't already on a web site priming us to look for them. How much of the skepticism is because you are on "alert" vs one's baseline levels of skepticism.
- tluyben2 5y agoTo me they always look like uncanny valley so my brain flags them as fishy. Everything deepfake feels like watching Gemini Man to me. But then again, I am pretty paranoid as I do think the media and web are not out there to provide me with quality; they want to achieve money and/or power and will try to manipulate me in order to do that so I would assume fake in case of doubt.