Which person is real?

#1

One of the two images is real, the other is generated by GAN (generative adversarial neural network)

I can get 9 out of ten right, but being stumped by even 1 is scary stuff.

1 Like

#2

Crap. I got 6/10.

I don’t like computers anymore.

OMG. What if I’ve been generated by a GAN and I don’t even know it? And that’s why I can’t tell the difference between real faces and generated faces? Mind blown.

0 Likes

#3

The more I play it, I increasingly get it wrong.

I have a few observation so far…

  1. Usually the one with a more complex background is the real person.
  2. Usually the one with accessories, hats, bling, earphone and so on is the real person.
  3. Usually the one with another person in the shot is the real person.
  4. Usually the ugly one is the real person.

With all that, I still get like 4 out of 10 wrong…

1 Like

#4

The distorted, chopped up backgrounds are definitely the biggest giveaway in the computer-generated images.

1 Like

#5

I got 7/10. I found I got it wrong when I second guessed myself.

1 Like

#6

I had the same experience as @hansioux. I got 9/10 right at first, but then my performance started getting worse and worse. Some of those fakes are pretty impressive in their level of quirky detail.

1 Like

#7

I get the feeling that we are helping to train the network…

2 Likes

#8

I was also thinking the whole thing may be an elaborate ruse by some AI intelligence to scare humans into thinking it’s more advanced than it really is.

1 Like

#9

If the AIs are already skilled enough to practice deception, we’re doomed.

Anybody seen “Ex Machina”? The story falls to bits at the end, but it’s an interesting premise.

1 Like

#10

First ten I got 6.

Second ten I got 10.

Rule is pick the image that looks like it’s the most fake.

So while the technology is very good, the dismay of fail here comes mainly from the test itself (the comparison presented), not from the tech.

0 Likes