This is like when you learned programming 20 years ago and realized that if you randomly generated a 300x300 pixel image you could get a nude picture. Exciting until you learned how to calculate the odds…
You know how progress goes in this industry. The first nude I ever downloaded was 320x200x16, dithered. All I had to do was stand back 5 feet, squint a li'l...
> one of the standard test image (for early machine learning maybe?) was a woman in a suggestive pose (from Playboy or similar magazine if I remember correctly)
I remember a body positivity blog let contributors provide nude photos, often mobile. To show how normal bodies look like. The engineer in me wondered “wait, what if…” so I used an online EXIF viewer and yep, it wasn’t stripped. A map pin right on their house! :o That was the worst mistake I’ve seen with these yet.
Downvoted! The title might have changed now, but the original one was completely indecipherable. As far as I could tell it was genuinely some sort of image recognition algorithm for naked buttocks.
With Stable Diffusion + controlnet, it is pretty damned easy to make nudes of people - and if they're wearing something tight-fitted, it's probably quite accurate as far as overall body shape.
Of course, you can make a model of someone with ~30 images and have them wearing and doing whatever you want.
On the flip side, that means if your real nudes are leaked you have plausible deniability. What might have been a career ending incident can now be dismissed as someone generating those images instead.
I’d like to believe that this is the future. Already with the rise of digitally native relationships nudes have become commonplace (even Jeff Bezos has sent some). Now with these widely accessible deep-fake generators any leaked nude photo can be chalked up to digital malfeasance!
reply