#1  
Old 02-14-2019, 08:59 PM
jasg is offline
Charter Member
 
Join Date: Nov 2001
Location: Upper left hand corner
Posts: 5,819

This person does not exist


An interesting web site - AI generated, photorealistic human portraits. Refresh the page to get a new image.

https://thispersondoesnotexist.com

Article here.
  #2  
Old 02-14-2019, 09:47 PM
JKellyMap's Avatar
JKellyMap is offline
Member
 
Join Date: Mar 2005
Location: Wisconsin
Posts: 9,294
Wow. That’s deep — really. Brings up all sorts of inchoate thoughts about identity and human individuality.

Last edited by JKellyMap; 02-14-2019 at 09:48 PM.
  #3  
Old 02-14-2019, 10:00 PM
burpo the wonder mutt's Avatar
burpo the wonder mutt is offline
Guest
 
Join Date: May 2012
Location: NE Florida
Posts: 23,154
2200 more and James Cameron can start work on "Titanic II."

ETA: My mistake, there might already be that many, if not more.

Last edited by burpo the wonder mutt; 02-14-2019 at 10:03 PM.
  #4  
Old 02-14-2019, 10:00 PM
Dag Otto is offline
Guest
 
Join Date: Sep 2003
Location: Las Cruces
Posts: 5,264
Seems that there are varying values for "photorealistic". The first one I saw was grotesque and make me think - Virtual Mr. Potato Head!
  #5  
Old 02-14-2019, 10:16 PM
Roderick Femm's Avatar
Roderick Femm is offline
Member
 
Join Date: Jul 2005
Location: On the cusp, also in SF
Posts: 6,895
I can't put my finger on why, but there seems to be a lot of androgyny going on there. Many of the photos seemed to be amalgams of male and female facial parts, and except for cues like long hair and jewelry (for women) and short hair and facial hair follicles (for men) I wouldn't be able to tell them apart.
  #6  
Old 02-14-2019, 10:16 PM
Tibby's Avatar
Tibby is offline
Member
 
Join Date: Oct 2005
Posts: 2,666
There's a photo of me on there!
  #7  
Old 02-14-2019, 10:50 PM
jasg is offline
Charter Member
 
Join Date: Nov 2001
Location: Upper left hand corner
Posts: 5,819
Quote:
Originally Posted by Tibby View Post
There's a photo of me on there!
I thought that was you.
__________________
It is useless to attempt to reason a man out of a thing he was never reasoned into.
~ Jonathan Swift (1667 - 1745)
  #8  
Old 02-14-2019, 11:09 PM
Kamino Neko's Avatar
Kamino Neko is offline
Guest
 
Join Date: Apr 1999
Location: Alternate 230
Posts: 15,183
Quote:
Originally Posted by Tibby View Post
There's a photo of me on there!
So, how are you dealing with not existing?
  #9  
Old 02-15-2019, 05:16 AM
JKellyMap's Avatar
JKellyMap is offline
Member
 
Join Date: Mar 2005
Location: Wisconsin
Posts: 9,294
Quote:
Originally Posted by Kamino Neko View Post
Jorge Luis Borges would have liked that post. And the site itself — a sort of Library of Babel of individuals (rather than books), where you “wander” desperately, pointlessly, hopelessly trying to find that someone you recognize, or that someone you love.

Last edited by JKellyMap; 02-15-2019 at 05:16 AM.
  #10  
Old 02-15-2019, 06:20 AM
kayaker's Avatar
kayaker is offline
Member
 
Join Date: Jul 2009
Location: Western Pennsylvania
Posts: 30,921
The first picture I saw I would have sworn was Michelle Obama. I had to open another tab and image search for a picture of her in order to realize it wasn't. I refreshed the page and there was a picture of some famous child actor, or so I thought.
  #11  
Old 02-15-2019, 12:52 PM
Darren Garrison's Avatar
Darren Garrison is offline
Guest
 
Join Date: Oct 2016
Posts: 9,766
Quote:
Originally Posted by Tibby View Post
There's a photo of me on there!

Tibby, or not Tibby?


The same system can generate other things, such as cats.
  #12  
Old 02-15-2019, 01:11 PM
Ann Hedonia's Avatar
Ann Hedonia is offline
Guest
 
Join Date: Apr 2009
Posts: 3,023
Quote:
Originally Posted by Dag Otto View Post
Seems that there are varying values for "photorealistic". The first one I saw was grotesque and make me think - Virtual Mr. Potato Head!
The skin texture looks odd on most of them, especially in the forehead area
  #13  
Old 02-15-2019, 01:28 PM
Maserschmidt's Avatar
Maserschmidt is offline
Guest
 
Join Date: May 2009
Location: New England
Posts: 5,720
The intersection of hair and glasses is sometimes a trainwreck.
  #14  
Old 02-15-2019, 09:03 PM
Chronos's Avatar
Chronos is offline
Charter Member
Moderator
 
Join Date: Jan 2000
Location: The Land of Cleves
Posts: 81,782
I find the misses more interesting than the hits. A lot of them seem to have a funhouse-mirror distorted image of the same face splattered against the edge of the picture. Shoulders and torso are sometimes scrambled, or absent. And one I saw had a pair of glasses propped up on top of her head, except that they were distorted in ways that glasses never would be.
  #15  
Old 02-16-2019, 01:05 AM
Peter Morris's Avatar
Peter Morris is online now
Charter Member
 
Join Date: Apr 2003
Location: The far canal
Posts: 12,298
Quote:
Originally Posted by Darren Garrison View Post
.... such as cats.
AAAARRRRGGGGHHHH

Last edited by engineer_comp_geek; 02-16-2019 at 07:33 AM. Reason: fixed link mangled by quote
  #16  
Old 02-16-2019, 04:02 AM
clairobscur is offline
Charter Member
 
Join Date: Aug 2001
Location: Paris
Posts: 17,489
It's interesting, because I've been doing 3D images for a time, and I see that I now notice a lot of things that would have escaped me before.

Anyway, that's an amusing link, but, whatever could be the method they use to generate portraits (they seem to be composite), most pictures have significant flaws, and a number have terrible flaws (like an eyebrow on the forehead, a bit of beard on one side of a "female" face, a significantly deformed chin, etc....) The most realistic tend to be those of young people, because of course a smooth skin is easier to imitate than an older one.

Of course, there are tons of pictures out there that are more realistic than this, the originality here being the generation of so many different faces.
__________________
S'en vai la memoria, e tornara pu.
  #17  
Old 02-16-2019, 08:32 AM
Tibby's Avatar
Tibby is offline
Member
 
Join Date: Oct 2005
Posts: 2,666
I chose a random photo from that site and added a little animation ... strictly in the interest of science, of course.

I may be spending a bit less time on the SDMB for a while.
  #18  
Old 02-16-2019, 11:25 AM
pool is offline
Guest
 
Join Date: Jul 2002
Location: Inside
Posts: 4,258
It seems like in some pictures the AI sort of peppers in stuff like wrinkles or skin blemishes in a weird way. Also I guess AI doesn't know about braces or tooth whitening because none of the people have straight teeth or bright white teeth they all seemed stain but in the same generic way, also all the people seem to have a midline discrepancy sort of like Tom Cruise, weird.

There definitely were some faces though that looked extremely realistic.
__________________
"You can do anything you set your mind to...But money helps"

Last edited by pool; 02-16-2019 at 11:26 AM.
  #19  
Old 02-16-2019, 11:33 AM
Defensive Indifference is offline
Guest
 
Join Date: Jul 2007
Location: St. Louis, MO
Posts: 6,709
Is there a page that explains how it's making the images?

There was one of a woman wearing glasses. The image even had the distortion behind the lenses, where the eyes are squished in, the way they are when you wear thick glasses for nearsightedness. That was a nice touch.
  #20  
Old 02-16-2019, 11:52 AM
DPRK is offline
Guest
 
Join Date: May 2016
Posts: 2,763
Quote:
Originally Posted by Defensive Indifference View Post
Is there a page that explains how it's making the images?
Here's the link. You can also download the software. It is done with a neural-network architecture they describe as a "style-based generator", which incorporates an intermediate "style" layer they can use to more easily get at features such as hairstyle, pose, face shape, eyeglasses, coloring, lighting, etc. Like other neural networks, it is initially trained on a sample set of images.
  #21  
Old 02-16-2019, 12:25 PM
Darren Garrison's Avatar
Darren Garrison is offline
Guest
 
Join Date: Oct 2016
Posts: 9,766
It is probably similar in principal (but more sophisticated in exectuion) to this. (I used to have a small executeable of something like this where you could use any image that you want, but I'm not seeing it now.)


(Also--a really practical spinoff from this kinda thing.)
  #22  
Old 02-16-2019, 12:35 PM
bibliophage is offline
Charter Member
Charter Member
 
Join Date: Apr 2000
Location: Maine
Posts: 10,570
Apparently black people who don't exist don't exist. I must have flipped through 50 pictures, and I didn't see one who looked black or sub-Saharan African. There were one or two who looked like they might be north African.
  #23  
Old 02-16-2019, 12:39 PM
Defensive Indifference is offline
Guest
 
Join Date: Jul 2007
Location: St. Louis, MO
Posts: 6,709
Quote:
Originally Posted by DPRK View Post
Here's the link. You can also download the software. It is done with a neural-network architecture they describe as a "style-based generator", which incorporates an intermediate "style" layer they can use to more easily get at features such as hairstyle, pose, face shape, eyeglasses, coloring, lighting, etc. Like other neural networks, it is initially trained on a sample set of images.
Cool, thanks.
  #24  
Old 02-16-2019, 12:51 PM
Peter Morris's Avatar
Peter Morris is online now
Charter Member
 
Join Date: Apr 2003
Location: The far canal
Posts: 12,298
Are the pictures generated on the fly when you click the link? Or is it a random selection from a library of previously generated images?
  #25  
Old 02-16-2019, 01:03 PM
DPRK is offline
Guest
 
Join Date: May 2016
Posts: 2,763
Quote:
Originally Posted by Darren Garrison View Post
It is probably similar in principal (but more sophisticated in exectuion) to this.
That does not seem similar. Both programs have involve pictures, but the approaches are otherwise completely different. Moreover, they are not even solving the same problem: one is an attempt to optimize a collection of triangles so that they resemble a given input image (e.g., Mona Lisa), while the other attempts to learn and imitate the class of images it sees (so trained on, say, cars it will output imaginary cars of its own design).

In your link, the program evolves chromosome data used to draw a picture using a genetic algorithm. Each individual is evaluated on how closely it reproduces the sample image. In the Nvidia paper, a collection of 70000 images (not one image) is used to train a neural network (there does not seem to be any evolutionary programming involved). The one thing in common is that there is evaluation of the performance of the network, but the performance is based on how well (quality and diversity) it reproduces an entire class of images (not a single image), and the quality also takes into account presence and absence of "features", not merely a pixel-to-pixel comparison.
  #26  
Old 02-17-2019, 12:14 AM
BigT's Avatar
BigT is offline
Guest
 
Join Date: Aug 2008
Location: "Hicksville", Ark.
Posts: 35,840
Quote:
Originally Posted by Peter Morris View Post
Are the pictures generated on the fly when you click the link? Or is it a random selection from a library of previously generated images?
The linked article in the OP says it produces a new picture every two seconds.

It seems to allow me to reload a bit faster than that, so I would guess it keeps at least a small cache to serve when the next picture isn't ready. Though it's possible it actually does put in a slight delay in loading if it isn't ready.

Last edited by BigT; 02-17-2019 at 12:16 AM.
  #27  
Old 02-17-2019, 12:37 AM
Gatopescado's Avatar
Gatopescado is offline
Guest
 
Join Date: Aug 2001
Location: on your last raw nerve
Posts: 21,435
I call 'Bullshit'.
  #28  
Old 02-17-2019, 12:55 AM
El_Kabong's Avatar
El_Kabong is offline
Member
 
Join Date: Aug 2000
Location: Smack Dab in the Middle
Posts: 15,229
Quote:
Originally Posted by Chronos View Post
I find the misses more interesting than the hits. A lot of them seem to have a funhouse-mirror distorted image of the same face splattered against the edge of the picture. Shoulders and torso are sometimes scrambled, or absent. And one I saw had a pair of glasses propped up on top of her head, except that they were distorted in ways that glasses never would be.
I went through about thirty iterations and while I didn't find anything wildly out of line like the above, there was a certain subtle...oddness to at least half the pictures. For example, one person was wearing eyeglasses that were slightly different between the left and right halves. Others had seemingly unlikely proportions among the upper, middle and lower portions of their faces, and I noticed a tendency toward androgeny as mentioned by a couple of other posters.

Still, interesting. Thanks to the OP for posting that.

Last edited by El_Kabong; 02-17-2019 at 12:55 AM.
  #29  
Old 02-17-2019, 01:24 AM
BigT's Avatar
BigT is offline
Guest
 
Join Date: Aug 2008
Location: "Hicksville", Ark.
Posts: 35,840
Quote:
Originally Posted by Gatopescado View Post
I call 'Bullshit'.
Seems right in line with Deepfakes to me. (Video is more complicated, but they also discard more unrealistic results.)
  #30  
Old 02-17-2019, 04:30 AM
Channing Idaho Banks's Avatar
Channing Idaho Banks is offline
Member
 
Join Date: Mar 2015
Location: beautiful Idaho
Posts: 2,661
, I looked at some of those images and there were elements that reminded me of Fred Savage.
  #31  
Old 02-17-2019, 05:10 AM
Half Man Half Wit's Avatar
Half Man Half Wit is offline
Guest
 
Join Date: Jun 2007
Posts: 6,517
This sort of thing, I think, serves better to illustrate the failings of AI, than its successes (while they are impressive). The errors are due to the AI fundamentally not knowing what it is that it's producing; to it, it's just a set of pixels sufficiently close to reference data, but not something like 'a face' or 'a cat'. It needs an additional model of do's and don't for cats and faces---it needs to somehow know that cats don't have giraffe-like necks, for which it needs to know what a neck is; that is, it needs some model of the components of a cat, what assemblage of parts makes a cat, and what those parts themselves are. I think this is really going to be the next big challenge in AI.

I can't help but be struck by the similarity to what psychologists call 'System I' versus 'System II'-thinking. System I is the sort of fast, heuristic, recognition-engine that seems to work like a neural network; System II then comes in to create an explanatory hypothesis, build a model of what's perceived and check whether that model is plausible, fits known models, and so on. So it seems like AIs lack System II-thinking to 'weed out' stuff that System I recognizes as being similar enough to some reference class, but which contain disqualifying elements only obvious upon checking the semantics of the images.
  #32  
Old 02-17-2019, 11:19 PM
Karen Lingel is offline
Straight Dope Science Advisory Board
 
Join Date: Jun 1999
Location: Union City, CA
Posts: 1,291
Most of the women are wearing unmatched earrings!

Unless this is some hip new fad about which I was not informed...
__________________
-k-
Karen Lingel, Physicist and Penguinist
  #33  
Old 02-18-2019, 06:31 PM
Mangosteen is offline
Guest
 
Join Date: Jul 2005
Location: Namche Bazaar
Posts: 2,476
Quote:
Originally Posted by bibliophage View Post
Apparently black people who don't exist don't exist. I must have flipped through 50 pictures, and I didn't see one who looked black or sub-Saharan African. There were one or two who looked like they might be north African.
The second and fifth images I got were of black people.
  #34  
Old 02-18-2019, 08:15 PM
darek is offline
Guest
 
Join Date: Feb 2019
Location: Albany, New York
Posts: 7
I don't know why, but neural network generated faces always creep the hell out of me, especially since they're getting more and more realistic.

Last edited by darek; 02-18-2019 at 08:16 PM. Reason: added a word
Reply

Bookmarks

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is Off
HTML code is Off

Forum Jump


All times are GMT -5. The time now is 05:11 AM.

Powered by vBulletin® Version 3.8.7
Copyright ©2000 - 2019, vBulletin Solutions, Inc.

Send questions for Cecil Adams to: cecil@straightdope.com

Send comments about this website to: webmaster@straightdope.com

Terms of Use / Privacy Policy

Advertise on the Straight Dope!
(Your direct line to thousands of the smartest, hippest people on the planet, plus a few total dipsticks.)

Copyright 2018 STM Reader, LLC.

 
Copyright © 2017