Nvidia’s latest offering has people around the world second-guessing everything they see online.
“Nvidia Corporation” is an American technology company based in California, that back in October developed artificial intelligence (AI) that can generate disturbingly realistic images of completely fake people. Now, they went one step further by creating AI that can generate fake videos as well.
The company’s creation can easily change day into night, winter into summer, house cats into cheetahs and much more. However, maybe one of the best features of the AI is the fact that it can achieve all of this with much less training than any other existing system.
The image translation AI uses a type of algorithm named GAN, or generative adversarial network. It’s a combination of two neural networks working against each other. One of the networks creates an image/video, while the second one critiques its work.
What makes this AI incredible is that it can convert a normal street into a snow-covered one, without ever actually seeing it. Usually a GAN collects significant amount of data in order to learn how to generate its own data. That means that the system needs to see more images of what a street would look like when it’s covered in snow, in order to generate an image on its own.
The masterminds behind Nvidia’s creation are the researchers Ming-Yu Liu, Thomas Breuel, and Jan Kautz.
Trusting Your Own Eyes
Lui told The Verge that there are several potential applications in which the AI could be adopted, but didn’t want to comment how quickly that would happen and on what extent.
“For example, it rarely rains in California, but we’d like our self-driving cars to operate properly when it rains. Now we can use our method to translate sunny California driving sequences to rainy ones in order to train our self-driving cars”, he explained.
Besides practical, this technology could also be very playful. Imagine being able to see how your future home would look like during different seasons of the year, or a potential wedding location covered in flowers during spring.
Unfortunately such technology could also be used in a bad way, and if it becomes widely adopted we will lose our ability to determine if something we see online is real or fake. Not to mention fake news becoming even more widespread on the internet, or not being able to use video evidences in court because of their irrelevance.
At the moment the capabilities of this AI are limited to just a few applications, so we have no way of telling how it will impact society once it gets into consumers’ hands, we can only hope for a proper regulation, because, clearly, the impact of AI is going to reach well beyond the job market.