Online content is created every second, and a huge part of that content consists of photographs. We take around 380 billion photos a year, and 300 million photos are uploaded to Facebook every day. But what makes those photos memorable? MIT engineers came up with an algorithm that predicts how memorable photos will be.
- Every two minutes, we take more pictures than the whole of humanity in the 1800s
- It is estimated that humanity has taken around 3.8 trillion photos
- 5,000 people were used as subjects for MIT’s algorithm
- The most popular photos on Flickr are of revolvers and women in bikinis
Scientists from the Computer Science and Artificial Intelligence Laboratory, or CSAIL, which you might remember won a robot from NASA on which to perform R&D, came up with an algorithm that can predict which photos will be more memorable.
This is a skill that even most humans lack, and even though the algorithm was implemented, it’s easier for it to actually predict the pictures people will remember than to actually explain why.
This is how it goes.
The scientists got a sample of 5,000 people scattered throughout the world, and employed them via Amazon’s Mechanical Turk service.
Each subject was shown hundreds of pictures, at speeds of 1 per 600 milliseconds, and they were then asked to note which ones they found memorable.
The researchers introduced the pictures into the algorithm, and told it which ones were memorable, and which ones weren’t. From here, the program managed to find patterns among the images, and is now 30% better at recognizing the remarkability of photos than any other algorithms out there.
However, it cannot really explain the reason why the pictures it selects are memorable, so the scientists still don’t know exactly how a remarkable photograph can be quantified.
Some are considering the applications in regards to social media. Using the algorithm, you would be able to ascertain which of the photos you took will get more likes, or even what photos to use on your online dating profile.
Others are considering the applications on a much broader scale. The algorithm could very well be used in marketing, making commercials more memorable, or it could even be used in the educational system, presenting memorable photos along with relevant text, so as to make learning easier.
The applications aren’t really endless for a program that
photos people will remember, but there is undoubtedly a broad spectrum of applications for the algorithm.