Condensing AI To Run On Mobile Phones is Facebook’s Latest Breakthrough


Facebook has shown over time that they are quite serious in integrating Artificial Intelligence in their operations. They used AI to help blind people use their social network and also incorporate AI into their app Moments in a very interesting way.

Today, Facebook gave a detailed roadmap of what they are planning to do with AI and it shows that they are pretty serious about it.

First, they explained on a technology that we have seen Prisma employ on their mobile phones. Facebook has something called style transfer, which is used to apply an artistic style of painting on each frame of a video. This usually requires the the individual frames to be sent to the data centres for processing and that will make the whole process slow.

Now, Facebook says that they have been able to shrink the AI model 100 times and came up with Caffe2Go.This is a neural learning platform that will be able to process these images and video on the mobile apps. This is reminiscent of what Prisma was able to do with their brilliant app on both iOS and Android where you can apply the neural net powered filters on your photos offline.

This is not the only AI venture Facebook is planning in their future. They will use it on VR, their Aquila program of providing internet connectivity and also will research more into it so that “computers will be able to learn, plan and reason like humans.”

It is ridiculous that we are in a time where we can have deep learning neural networks in our phones which are becoming more powerful every minute. There will be a time when these will be normal and will usher the age of extremely personalized feeds for everyone.