Nvidia continues to pave a path towards the end of reality as we know it. The company recently made an AI capable of creating images of people out of thin air. Now, it’s made one that can change the weather, turn day into night, and change a leopard’s spots.
AI researchers for the company developed an unsupervised learning method for computers which allows for sweeping changes to video content it’s fed. By using the new method, they were able to produce startling results.
Previous methods required massive amounts of data by comparison. And the issue was made more difficult when trying to train the machines to find their own patterns, the basis of unsupervised learning.
They considered the problem similar to teaching a machine to colorizing a photo, according to the researchers’ white paper:
For example, super-resolution can be considered as a problem of mapping a low-resolution image to a corresponding high-resolution image; colorization can be considered as a problem of mapping a gray-scale image to a corresponding color image.
Today the machines are mapping sunny days to rainy ones and creating the equivalent of a “snow plow” filter for videos. It’s a little unnerving to imagine what these things will be capable of in a decade.
Computers around the world are being trained in the art of deception. Nvidia’s recent work in the field of artificial intelligence shows the leaps and bounds by which AI has attained a clear ability to imitate reality to a startling degree.
Wired’s Oli Franklin-Wallis might have the right idea when it comes to the real worries for our AI filled future:
We may not be at “Matrix” levels of immersion yet, but just like running a film franchise into the ground, these things take time.