This Person Does Not Exist.
The link below uses AI to generate human faces that don't exist. Follow the link and just refresh the page to generate a new face!
The next 10 years are going to destroy models, actors, and criminal forensics as AI and Machine Learning have been getting exponentially better at creating human faces. Artificially created, photo-realistic pictures and videos are going to fuck with what people perceive as real or fake.
Offsite Linkby Anonymous | reply 22 | August 1, 2020 4:29 AM
|
Malformed faces sometimes show up in the periphery of the images. It’s fucking creepy.
by Anonymous | reply 1 | August 1, 2020 2:34 AM
|
Yea, the algorithms are not perfect.... yet, but they are constantly improving. The usual tell tale signs at this point that it is fake are the periphery of the image that you mention, eyes, hairlines, and teeth.
Again, if this is what they can do now, imagine it in 10 years.
by Anonymous | reply 2 | August 1, 2020 2:39 AM
|
This is cool, OP. thanks!
by Anonymous | reply 4 | August 1, 2020 2:55 AM
|
They look more real than some famous like certain CEOs or presidents
by Anonymous | reply 5 | August 1, 2020 3:01 AM
|
How do they know those faces don't exist? Have they seen all the billions of faces that exist?
by Anonymous | reply 6 | August 1, 2020 3:02 AM
|
Well, R6 even though the 7.8 billion people on the planet SEEMS like a large number, the mathematical permutations and combinations that can be done with a person's face based on human facial features is a trillion trillion trillion times more.
When you get into large numbers like that, coming across an identical face, even with 7.8 billion real people, is statistically unlikely.
Offsite Linkby Anonymous | reply 7 | August 1, 2020 3:08 AM
|
The image is not real. Period, if someone looks like it, that's another thing.
by Anonymous | reply 8 | August 1, 2020 3:40 AM
|
Whoa. Some of them look very real. I’m scared.
by Anonymous | reply 9 | August 1, 2020 3:51 AM
|
R1, like this one!
I think it does a better job with female faces.
Offsite Linkby Anonymous | reply 10 | August 1, 2020 3:55 AM
|
Yes, it seems biased to generate a lot of female and children's faces not sure why tho....
R9 you should be scared!
by Anonymous | reply 11 | August 1, 2020 4:00 AM
|
And some dissolving eyeglasses.
Offsite Linkby Anonymous | reply 13 | August 1, 2020 4:04 AM
|
Oh my god, I could do this all night.
Offsite Linkby Anonymous | reply 14 | August 1, 2020 4:11 AM
|
Just one more. This one is freaking me out, man.
Offsite Linkby Anonymous | reply 15 | August 1, 2020 4:12 AM
|
So what I’m gathering is that the faces are pretty well symmetrical and regular. There don’t seem to be any real irregularities that I see in people I know. Like a long philtrum or a scar on the eyebrow (it’s very common!) or a hooked nose.
But the faces are mostly believable. The hair and some of the coloring. I’m trying to poke holes in this thing, because I’m actually alarmed at how real some of them look.
by Anonymous | reply 17 | August 1, 2020 4:16 AM
|
it only generates ugly people? If it can't make hot ones what's the point?
by Anonymous | reply 18 | August 1, 2020 4:16 AM
|
As the kids say “this thing is cursed!”
by Anonymous | reply 20 | August 1, 2020 4:19 AM
|
Haha I didn't even see the cat ones! But this is a common problem for machine learning algorithms. Shape and colors that look very similar confuse the algorithm.
A common example used to test the robustness of the algorithms is whether they can tell the difference between a dog or bagel.
Offsite Linkby Anonymous | reply 22 | August 1, 2020 4:29 AM
|