Newspaper article The Christian Science Monitor

How an MIT Algorithm Can Make Your Selfies More Memorable

Newspaper article The Christian Science Monitor

How an MIT Algorithm Can Make Your Selfies More Memorable

Article excerpt

A tantalizing development at the Massachusetts Institute of Technology could help boost the popularity ratings of selfie takers and online daters.

Scientists at the institute's Computer Science and Artificial Intelligence Laboratory (CSAIL) have taught computers to predict, with near perfect precision, which photos of faces, nature scenes, or other objects people are most likely to remember.

Though similar predictive algorithms in the field of machine learning already allow computers to predict information by automatically completing phrases people type into Google search or by recognizing people to tag in photos on Facebook, scientists have not until now been able to use these tools to teach computers to predict what will be memorable to people, a skill that even humans themselves lack, Aditya Khosla, a PhD student in computer science at MIT, told The Christian Science Monitor.

For selfies and other portraits, this development means an app might one day tell people which one from a group of photos is likely to get the most likes on Instagram, or to attract more suitors on a dating site. The team in 2013 developed an algorithm that can slightly modify pictures of faces to give them a more memorable flair.

"While deep-learning has propelled much progress in object recognition and scene understanding, predicting human memory has often been viewed as a higher-level cognitive process that computer scientists will never be able to tackle," said CSAIL principal research scientist Aude Oliva in an announcement.

"Well, we can, and we did!" said Dr. Oliva, a co-author of a paper on the topic with Mr. Khosla and other MIT researchers that was presented this week at the International Conference on Computer Vision in Chile.

To test its new algorithm, the CSAIL team hired 5,000 people from across the globe through a crowdsourcing site called Mechanical Turk, an Amazon company the connects developers with millions of people who are willing to perform simple tasks for small payments.

The scientists showed their workers hundreds of images, each one for 600 milliseconds, and recorded which ones were memorable, meaning that people were able to identify the image after seeing it once. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.