Artificial intelligence: What can’t it do? Will it solve the hunger crisis? Will it cure diseases? Will it foster world peace and understanding among fellow human beings of different creeds and cultures? Will it accelerate humanity’s dark slide into the depths of odiousness and depravity? Maybe!
But will AI make it so that your eyes seem open instead of closed? Yes!
When they’re not figuring out how to target your most deep-seated insecurities and serve you one of ten different mail-order mattress ads, the Wizards of Menlo Park over at Facebook are hard at work developing AI systems to cover closed eyes with realistic open ones in pictures. In the dryly named “Eye In-Painting With Exemplar Generative Adversarial Networks,” Brian Dolhansky and Cristian Canton Ferrer outline how they’ve created a better way to deal with accidentally blinking during a selfie.
Basically, the system is trained on a bunch of pictures of a subject with their eyes open to figure out attributes like shape and color, and then it “paints” similar eyes onto photos where your eyes are closed. The examples shown in the paper are, if nothing else, convincing facsimiles. Facebook has no plans to introduce the system to its audience; it’s merely research right now.
If you do happen to blink during a photo, you could, I dunno, just take the picture again. Or you could do this whole thing. Up to you. The term for this type of technological development is solutionism, coming up with a complex solve for a problem that is hardly a problem. That Facebook — a company currently struggling with a serious perception problem as a service full of inauthentic posts, sock puppetry, and fake news — would pursue such tools is … curious.