Nvidia is a well-known technology company especially among gamers as it makes the popular GeForce graphics processing unit (GPU), as well as System-On-Chip (SOC) for mobile computing and the automotive industry.
At the ongoing CES (Consumer Electronics Show) 2017 in Las Vegas, Nvidia founder and CEO, Jen-Hsun Huang introduced the company’s new supercomputer, named Xavier, based on its Drive PX2 supercomputer and called it “your most personal robot”. According to Huang, Nvidia has been developing self-driving technology for the past ten years and Xavier represents the latest and greatest from the Santa Clara-based company.
Xavier is an SOC with an octa-core ARM 64 processor and Volta GPU capable of 30,000 trillion operations per second (TOPS) enabling this artificial intelligence (AI) system to determine the vehicle’s location, process data from an array of sensors and cameras and make the appropriate intelligent driving decisions. This AI isn’t just reliant on a set of programed instructions but is capable of deep learning i.e. it ‘remembers’ and learns from observation and experience, allowing it to adapt to real world scenarios. Does that sound like a self-aware Skynet?
Nvidia is collaborating with various automakers and suppliers such as Audi, Here, Bosch and ZF to bring its self-driving technology to life. Some industry analysts predict automakers will be adopting this technology over the next ten years.
There is an interest to make cars autonomous to reduce accident rates and cut traffic time. Currently, self-driving aids like the Tesla Autopilot is considered SAE Level 3 which means it has conditional self-driving automation and many automakers are aiming for the ultimate Level 5 which is full automation. It is said that Tesla is already building Level 5 capability into their new Models S, 3 and X but it will not be enabled until comprehensive tests are completed and validated.
Check out Nvidia’s self-driving demo with its Lincoln test car, nicknamed BB8:
This is an earlier demo of the car’s deep learning capacity: