Future Tech
Deepfaking AIs are inventing faces and bringing the dead to life
It's hard to believe: refresh, refresh, refresh - a different person every time you reload the page. Are all these people okay with their faces being online in high resolution? Or is this some strange database of wanted criminals?
None of these faces are real, a prompt soon tells you - even though you could have sworn otherwise.
You're on the website ThisPersonDoesNotExist.com, where developer Phil Wang showcases fake faces - computer-generated images of people who simply do not exist.
The photos are known as deepfakes, and this is by far not the only website of its kind. Several other projects are bringing faces to life - both real and imaginary.
There's the smiling visages of Altered Qualia, whose eyes will eerily follow your cursor, with plenty more examples on alteredqualia.com.
Some people are even bringing deceased loved ones and famous personalities to life using a similar technology. A glance at the #DeepNostalgia posts being shared on Twitter right now shows countless videos of animated faces in black and white, from Anne Frank to late grandparents.
Many of them come from a genealogy portal that has recently started offering its customers the opportunity to bring their own uploaded portraits to life.
Some videos show people shocked at the sight of a photo of a dead parent being reanimated with a moving head, blinking eyes and a smile or two. It's called Deep Nostalgia, and it's surprisingly realistic, even with faces from paintings or sculptures.
Meanwhile over on ThisPersonDoesNotExist.com, the deepfaked, imagined faces come from two interconnected neural networks that are continuously trained with photos of real people to create these impressively realistic portraits.
One neural network calculates entirely new faces from all the possible human features it has memorised during learning, from face shapes to skin characteristics and eye colours to hairstyles.
The other neural network then evaluates the result with its own empirical values. If the fake face passes the "reality check", it's published. Otherwise, the first network has to rework everything.
More than 90% of the results are so realistic that if you didn't know how the portrait came about, you'd have no doubts about its authenticity.
But here and there, you'll notice a glitchy face, which is somehow reassuring. There might be something wrong with the symmetry of the face, the peak of a baseball cap is growing out of someone's forehead or there are texture errors on the skin.
- dpa
More articles on Future Tech
UK semi industry exposed to supply chain risk, China state ownership
Created by Tan KW | Aug 13, 2024
AI chatbots amplify creation of false memories, boffins reckon – or do they?
Created by Tan KW | Aug 13, 2024
Dispossessor ransomware group shut down by US, European authorities
Created by Tan KW | Aug 13, 2024
Sri Lanka regulator grants licence to Musk's Starlink for satellite broadband
Created by Tan KW | Aug 13, 2024
DEF CON badge disagreement gets physical as firmware dev removed from event stage
Created by Tan KW | Aug 13, 2024
Discussions
Be the first to like this. Showing 0 of 0 comments
Post a Comment
Featured Posts
Latest Videos
Apps
Top Articles
1
THE INVESTMENT APPROACH OF CALVIN TAN
2
M+ Online Research Articles
Vetece Holdings Berhad (VTC) - Driving Growth In Tandem With Its Technology Partners
3
save malaysia!
6
7
save malaysia!
8
Good Articles to Share
#
Stock
Score
Daily Stocks
Stock Name
Last
Change
Volume
Stock Name
Last
Change
Volume
Stock Name
Last
Change
Volume
MQ Trading Signals
Stock
Time
Signal
Duration
Stock
Time
Signal
Duration
Featured Advertisers / Partners
Ride The Bull Short The Bear
CS Tan
4.9 / 5.0
This book is the result of the author's many years of experience and observation throughout his 26 years in the stockbroking industry. It was written for general public to learn to invest based on facts and not on fantasies or hearsay....
