Nvidia's self-driving car test showcases company's shift to AI solutions

Nvidia began testing its autonomous vehicle tech on California streets last week, showcasing big investments that the firm has made in AI and deep learning.

nvidia.jpg
Image: Nvidia

Nvidia officially began testing its driverless car technology on public roads in California last week, shortly after receiving a permit from the California Department of Motor Vehicles to do so. The test further showcases Nvidia's move from graphics card manufacturer to a provider of AI solutions.

One of the company's first big efforts in automotive came when it announced its connected car platform, the Drive PX, at the 2015 CES. The following year, Nvidia announced the second iteration focused on the "very hard problems" of autonomous driving.

Additionally, Nvidia revealed its Xavier SoC to act as a control center for autonomous vehicles in the future. The firm also recently partnered with Baidu to develop a mapping system for autonomous vehicles as well.

SEE: Our autonomous future: How driverless cars will be the first robots we learn to trust

A lot of Nvidia's original testing for self-driving cars began at its New Jersey office. Using a form of deep learning called a convolutional neural network (CNN), the Nvidia team taught the car to steer itself. An Nvidia blog post said that the CNN was trained using a "time-stamped video from a front-facing camera in the car synced with the steering wheel angle applied by the human driver."

Instead of hardcoding certain "if, then" statements into the network, the Nvidia team simply taught the CNN to watch and learn from the human driver. As such, the post said, Nvida "never explicitly trained the CNN to detect road outlines," rather it taught the system to learn the "rules of engagement between vehicle and road."

So, you're probably wondering what Nvidia's background in graphics has to do with AI. The issue has to do with parallelization. GPUs do well with parallel computing, and many AI, machine learning, and deep learning problems fall under the category of parallel problems. Because of this, GPUs can handle these problems much faster and more efficiently.

Nvidia's expertise making GPUs for industries like gaming provided the perfect segue to create AI and deep learning solutions. However, the company isn't only dealing with autonomous vehicles. In November, Nvidia partnered with IBM to build a software toolkit that they call the "fastest deep learning enterprise solution" in the world.

The 3 big takeaways for TechRepublic readers

  1. Nvidia recently tested its self-driving car solution on California roads after receiving its permit from the DMV.
  2. Nvidia's expertise in GPU manufacturing made it a strong candidate to begin building out solutions for AI, machine learning, and deep learning.
  3. Nvidia is also working on AI toolkits, like one it built with IBM that it called the "fastest deep learning enterprise solution" in the world.

Also see