Crazy Consumer Electronics World: CES 2015: Nvidia sees a future of self-aware cars, and it's got some ideas on how to get there

CES is always about discussing what's possible, even it's not a product that'll be on the market this year, the next, or even the year after that.

Nvidia is following in the footsteps of many, discussing self-aware cars and the tools it's using to achieve such a thing during its CES 2015 press conference.

The key to creating self-aware (and eventually self-driving) cars, said CEO Jen-Hsun Huang, is heavily updatable software. Cameras, he said, are key to developing autonomous roadsters, and for a car to become truly self-aware, every camera needs to be connected to a central super computer.

Luckily, Nvidia took it upon itself to develop the Drive PX, a 2.3 Teraflops auto-pilot car computer equipped with two Tegra X1 processors and 12 camera inputs.

Deep learning

The Drive PX introduces two new technologies: Deep Neural Network Computer Vision and Surround Vision.

Deep learning is not a new concept, though its development is still in the nascent stages. Neural networks that pass through GPUs essentially recognize what's going on in the world around them. According to Huang, deep learning and neural nets help a car recognize images, like street signs or pedestrians, on its own.

PX system

This, Huang said, is a computationally intensive task, so a heavy duty CPU is needed to complete it. Being able to recognize and understand context is vital for a self-aware vehicle, Huang said, and it can learn to classify objects overtime.

What will power this energy intensive neural network? The 256 GPU core Tegra X1, of course.