(Correct Nvidia executive’s title in last paragraph)
Stephen Nellis
SAN FRANCISCO, Dec 1 (Reuters) – Nvidia on Monday released new open source software aimed at speeding the development of self-driving cars by leveraging some of the latest “inference” techniques in artificial intelligence.
Nvidia has become the world’s most valuable company as its chips have become central to the development of artificial intelligence. But the company also has an extensive software research arm and releases open source artificial intelligence code that other companies, such as Palantir Technologies, can adopt.
On Monday, Nvidia unveiled Alpamayo-R1 for self-driving cars. The software is a so-called “vision-speech-action” artificial intelligence model, which means the self-driving car translates what its set of sensors see on the road into a description using natural language.
Alpamayo’s breakthrough is that it talks to itself as it plots its path through the world.
For example, if a car sees a bike path, it will notice that it saw that path and is adjusting its course.
Most previous self-driving car software has been limited in the way it interprets the specific paths a car takes, making it difficult for engineers to understand what needs to be fixed to make the car safer.
Katie Washabaugh, product marketing manager for autonomous vehicle simulations, told Reuters: “One of the whole motivations for opening up this technology is to allow developers and researchers to … understand how these models work so that we as an industry can come up with standard ways to evaluate how they work.”
(Reporting by Stephen Nellis in San Francisco; Editing by Jan Harvey)
