Social networks host billions of profiles, each represented by a main image. That image may be real or not, but among all that digital noise, some users want their chosen photo to be memorable. How can you determine if a profile picture is memorable or not? The folks at MIT offer a demo called LaMem, in which an algorithm processes the image and assigns it a score.
It has been over a year since I last changed my profile picture on Facebook. On the one hand, I'm not much of a photo person, and on the other, I don't consider it that important. However, those who live and breathe for social media (which, by the way, are not limited to Facebook) need to have the ideal profile pictures... emphasis on the plural. The traditional way to evaluate the quality of such photos is to ask others for their opinion, but now there is another alternative: let an algorithm act as the judge.
The intervention of artificial intelligence in social networks is one of the worst-kept secrets of the industry. Entire platforms work behind the scenes to keep users on their sites longer, and of course, to show them more relevant advertising. What we have here today is called LaMem or “Large-Scale Image Memorability,”. Its goal is to objectively measure human memory through scores assigned to different profile images. The scale goes from 0 to 1, so if a photo gets a score of 0.88, it has an 88 percent chance of being remembered within the next hundred seconds.
A hundred seconds may seem insignificant for some tasks, but in the whirlwind of the Web, remembering something for almost two minutes gives it significant merit. Those interested in virtual reality developments can download a copy of LaMem (pre-trained with Caffe), for educational and/or non-profit purposes only. To close, I decided to use the profile photo that appears at the bottom of each article, and LaMem gave it a score of 0.81. I think the algorithm is being more generous than it should be...
