Okay here's the outline. With 15+ decent quality pictures of your face and upper-body I can train a model against a bigger model trained on billions of angles of human bodies, how to replicate your personhood into almost any situation. It's so real it's _too_ real. That's the only giveaway right now
good thing I'm a wolf
I see my plan to burn everything down and go live in the woods is a solid choice.
Good thing I’m a vampire.
how about... an old time celebrity who is long gone, and some archival photos, and set them in a modern tech environment?
The thing that I find most concerning is that the technical barrier to entry is falling *very* fast. We're going from 'techie hobbiests' to 'can follow along step-by-step' at a rate of knots.
There will come a time when someone hacks a Vision Pro and walks into the wearer’s view wearing the digital skin of their loved ones.
Well this is terrifying
Believe me I get what you’re saying but per se “inpainting” AI art models will always make subtle mistakes of perspective foreshortening, impossible body postures and positioning because, by their nature, they are not depicting a ray traced virtual representation of reality.
It occurs to me that if humans are clever as advertised then, like our descendants in a Frank Herbert novel we will develop distinct one-time pad sign and gesture languages that denote and encode an “authentic” photo or video.
Sir, please put down the Vortex of Suffering and quietly step back from it.