Started by several graduates from Imperial College London, Magic Pony uses large neural networks to process visual information, and it specializes in using partly unsupervised learning to conjure up images based on previous training. In other words, after feeding on lots of video footage, its algorithms can automatically fill in a patchy video feed, or increase the resolution of a pixelated video game image.
The results aren’t always perfect, because the system has no idea what it is actually creating. But usually they are good enough to fool the human eye. And so Twitter’s interest is unsurprising, given how much it is investing in live video, as well as live video processing, which requires a huge expenditure on data storage and computing power. Magic Pony’s technology might allow it to reduce the cost of streaming live video significantly.
I was lucky enough to meet with the team from Magic Pony at Nvidia’s developer conference this March, where they were demoing their technology. I was impressed by their expertise in machine learning, and their demos—especially in which they turned one of Vincent Van Gogh’s famous self-portraits into a photorealistic image of the impressionist master.
And, in case you were wondering: the company’s somewhat bizarre name came from a meeting at which an investor said the technology seemed almost too good to be true.