Human can now figure what is likely to happen next based on a image; with little or no effort.

A team of researchers are working on an artificial intelligence that can predict what will happen in the immediate future based on an image. An Al that animate a still image to what is likely to happen in the Future. The algorithm can also produce video on what is likely to happen in next scene.


HOW IT WORKS

Previous systems developed to create predictive videos have worked on a frame-by-frame basis, while MIT's system generates all 32 frames simultaneously.

The frame-by-frame method creates a larger margin of error, say the researchers.

Once the algorithm was able to generate videos, the researchers refined it using a deep-learning method called called 'adversarial learning'.

This involves using two neural networks compete against each other.

In this case, one generating videos, and one determining the difference between the real and the generated videos.

The idea was for the video-generating network to eventually be able to fool the other network into thinking that new videos are real. read more @ teazhacks..com