Twitter bought a Magic Pony neural network to sharpen images and videos

150 million dollars for "obvious technology"

Just a couple of months ago, Geektimes told about the little-known English startup Magic Pony Technology, which was founded by engineer Rob Bishop - one of the creators of the Raspberry Pi mini-computer.

The startup has developed a revolutionary image modeling technology that significantly improves the resolution of photos and video in real time. A trained neural network does not just interpolate pixels, but adds the missing details. The developers said that it is possible to automatically generate elements in this way, for example, for realistic virtual worlds.
')
In the comments on GT, they agreed that "this method is quite obvious, and its patent is almost patent trolling." Moreover, the Sadler user reported that two years ago he had already implemented such technology on autoencoders ( an example of the work of a neural network trained by Sadler ).

Sadler will be interested to learn that Twitter today announced the purchase of startup Magic Pony Technology. According to rumors , the amount of the transaction is $ 150 million. Not so little for the obvious technology.

According to the MIT Technology Review publication, when training a neural network, researchers generated low-quality images from high-resolution images - and submitted these pairs of images to the neural network's input. As a result, the neural network gradually learned to detect patterns and perform the reverse operation. The system does not require manual marking.


The image on the right is generated based on the sample on the left.

The developers claimed that their program is fast and efficient compared to competing technologies.

Twitter CEO and co-founder Jack Dorsey says machine learning is increasingly being used as the key technology for all Twitter systems. Trained neural networks are used to select the most relevant to users of news, stories and events. Therefore, the purchase of a startup that is engaged in development in the field of AI? Looks quite logical for Twitter.

The company can use the intellectual potential of the employees of Magic Pony Technology for their own needs. And of course, Twitter can somehow apply the Magic Pony Technology neural network itself, which enhances the sharpness of photos and videos.

An example of the operation of the Magic Pony Technology neural network is shown in a short video. On the left - the original video stream of the game, on the right - improved graphics.



In the future, technology can be implemented on personal computers and mobile phones, including web browsers, video players, computer game engines, and so on. Here, the program will effectively improve the quality of photos taken in poor lighting conditions or with low resolution matrix, as well as improve the quality of the video. In addition, such a neural network can be applied in image and video compression formats. However, only in proprietary formats (the technology is protected by 20 patents).

The staff of Magic Pony will join the Twitter Cortex team, which deals with artificial intelligence applications on Twitter. The staff includes 11 doctors of science in the field of computer vision, machine learning and high-performance computing, so this is a valuable acquisition. Perhaps Twitter company needs not so much a neural network as Magic Pony employees themselves.

For British AI developers, this is already the second major achievement in this technical area. In January 2014, Google acquired another London-based startup called DeepMind for half a billion dollars. That deal was a trigger for the real AI boom.

"Welcome to our herd, the Magic Pony!", - these words welcomed the new employees Jack Dorsey.