Submitted 4 months ago by BrikoX@lemmy.zip to technology@lemmy.zip
Researchers use technique to quantify eyeball reflections that often reveal deepfake images.
Simple, but quite clever.
Theoretically they could finetune for consistent eye reflections, but even then I suspect image models wouldn’t get it right. Physical correctness is not really something they actually do.
Just wait until Captcha starts asking this for a question.
This might be one case where I’d be okay with keeping the methods these tools use as a trade secret for at least a little while so as to slow the arms race. The longer we’re able to detect fake images the the better.
AI Engineer: So, more glints in the eyeballs?
ogmios@sh.itjust.works 4 months ago
Breaking: AI has discovered new methods to circumvent the latest detection techniques.