Email
Password
Remember meForgot password?
    Log in with Twitter

article imageOp-Ed: Fake faces by AI — A fraudulent fad and not smart

By Paul Wallis     Nov 21, 2020 in Technology
Sydney - AI-generated fake faces look very real. There are any number of possible faces, of all ethnicities. They can be used for fraud. They’re on the market for a few bucks. They’re also typical techno-hype in many ways. Pretty dumb, really.
The fake faces can be any age, race, with or without specific distinguishing features. The joke is that they’re a bit too good. Apparently they’re also now used by fake people to peddle conspiracy theories, fraud, and other wholesome early 21st century hobbies.
The New York Times did an explanatory article with many examples, and then proceeded dutifully to point out the flaws. The article is well worth a look, but I want to focus on some real ironies.
What’s in a face? A farce by any other name...
If you check out the fake faces on those links, you’ll notice a few extra things:
As The New York Times pointed out:
The eye spacing is all the same. That’s fixable, but it’s symptomatic of fake faces in general. That was the first thing I noticed, even before reading it in the NYT piece.
• Facial angles are always the same. The angles are usually full-face or three-quarter face, either side.
• Everyone is looking straight at the camera. It doesn’t look natural after a while. Very few people, except some excellent actresses, bother to look straight at a camera unless specifically told they should.
• There are very few people over the age of 30 without a few basic facial blemishes.
• The two sides of a real face aren’t identical. The fake faces are.
• The fake faces are ridiculously healthy. Most faces have transitory signs of minor conditions, etc.
• The skin on the fakes is flawless. Doesn’t happen in real life unless you’re actually taking Maybelline intravenously for decades. (Don’t.)
• The teeth are very similar. Also not too authentic.
• Fake hair doesn’t have split ends, dandruff, white or off-colour hair.
• The eyes lack distinguishing retinas. Wrong.
• Noses are characterised and resized, but the shape is much the same.
• These things wouldn’t survive a biometrics scan for long. Standardised faces would be very conspicuous in having metrics much too similar, particularly in ratios.
To be strictly fair to the “art form”, it faces the same problem portrait painters have had for millennia – Looking authentic and identifiable.
Who’s doing this?
I checked out the website cited by The New York Times, called Generated Photos.com. I got a domain name broker service screen. So the fake faces are generated by a fake domain? Or it disappeared soon after the NYT article? Uh-huh…
POSTSCRIPT: i have since been advised that this is the correct link for Generated Photos.com
Is it legal? Probably not, and never will be. It’s also pretty stupid.
There are any number of laws worldwide which say any sort of identity fraud is fraud. That does include fake images. The dumbest of the dumb may think they’re clever enough for a quick scam, but there’s a problem with that, too – Fake faces are documented evidence of fraud. If you’re associated somehow with a fake face, your credibility suffers.
Consider this – AI makes these faces. AI can also deconstruct and analyse them. Safe? No. Stupid? Yes.
This opinion article was written by an independent writer. The opinions and views expressed herein are those of the author and are not necessarily intended to reflect those of DigitalJournal.com
More about fake faces, AI generated faces, Artificial intelligence, fake faces and fraud
 
Latest News
Top News