Facebook will do whatever it can take to motivate you to share more content. Even, it turns out, with artificial intelligence to recover photos wasted by accidental blink. In a research paper published this week, a pair of Facebook engineers describe a brand new method which uses machine learning to retouch closed eyes in photos. The way is merely in the research stage at this time, although there is no warranty Facebook will introduce such a feature in its web site or applications, which does seem a strong potential.

As the researchers note, there are already plenty of photo retouching tools which could remove red eyes or smooth out flaws, so why not one which solves the issue of accidental jelqing? . The system works by utilizing a machine learning method called generative adversarial network or GAN. That is a technique which has proved especially adept at creating imagery, and has already been put to work producing fake celebrities, change the weather in the video, as well as designing clothes. Flashing is no challenge at all. For the system to successfully edit the photo it’s to first learn what an individual’s eyes typically look like.

This means training it on images when they are not blinking. From this, the software gets an idea of their eye shape and colour, and may then fill that info in for blinking images, a procedure referred to as in painting. , It is not an ideal solution, and the AI struggles to edit glasses, long fringes, and confronts at extreme angles. But when it works, it works, making incredibly realistic fakes such as those you see above. And AI tweaks such as this are only going to become more common. Software firms like Adobe and Pixelmator are working on several machine learning powered features that allow anyone to edit images with little technical ability.

From the not very distant future, artificial intelligence will do not only tiny changes, but generate fully new pictures and video. This may definitely create some problems for society, because it becomes easier for anybody to imitation visual evidence. However, it’s short term challenges too, including for Facebook. In the end, the business frequently touts its website as a house for authentic expression, it will give folks AI tools to modify their appearance? . And if yes, where will it draw the line? . Editing a blink seems pretty benign, but more considerable changes may be problematic.