3 ms·
I was completely expecting to be impressed with yet another demonstration of this technology taking a step forward. I got 31/32, and the only one I got wrong w
by apetresc 5y ago
I was completely expecting to be impressed with yet another demonstration of this technology taking a step forward.
I got 31/32, and the only one I got wrong was a text statement that I don't really understand the point of including. None of them were remotely difficult except for one.
- a_wild_dandan 5y agoEvery clip's authenticity was instantly, blatantly obvious. This survey could've been done in 60 seconds, sans the `x` seconds of required viewing before submitting an answer. I'm likewise baffled by the inclusion of text samples. It has nothing to do with deep fakes. Perhaps it's a control? I just skipped them, defaulting to the 50% skepticism slider option. Others in this thread report similar results. I wonder if Hacker News users have a preternatural aptitude for spotting inauthentic content owing to skewed technological competence, or if that's hubris -- simple primal readings of facial expressions/inflections expose the fakes. I'd bet on the latter option. Our lizard brains are keen to detect that uncanny valley. If this recording collection is any indication, duplicity is a few years down the road...