Nvidia touts new generative AI auto partnerships

SAN JOSE, California — Nvidia computing architecture is underpinning several new in-vehicle experiences, including advanced voice assistants, the company announced at GTC, a global AI conference it is hosting here.

Cerence, Wayve, Li Auto, Nio, Geely, SoundHound AI, Lenovo and other companies are designing user interfaces that incorporate generative AI, the type of intelligence behind chatbots such as ChatGPT. Nvidia's hardware and software enable this content in part through large language models, which learn to recognize data patterns to train tools including voice assistants.

Automotive AI is "enabling rapid content generation, all types of inputs and outputs, text, images, video sound animation," said Danny Shapiro, Nvidia's vice president of automotive. These generative AI workloads "are able to unlock all kinds of new possibilities, interfaces and applications."

Nvidia's new partnerships suggest the automotive industry is ramping up its use of artificial intelligence beyond the more complex applications of autonomy and advanced driver-assist tech. Companies are employing AI to improve interactions with the vehicle by beefing up voice assistants and other responses.

Shapiro said Nvidia has addressed the need for "increased computing horsepower" to handle sensor processing and running large language models. Companies are using vast amounts of information for training, including sentences written by humans, to enable a more natural response to queries from drivers.

AI tools in the vehicle can improve safety, convenience and entertainment, Shapiro said. The technology uses "basic speech recognition, command execution" and deep learning "to understand and process very complex sentences and instructions inside the car," he said.

Wayve, for example, is using Nvidia's platforms to support a vehicle "commentator" that uses natural language to explain the actions of the driver-assist technology to the driver and passengers.

Li Auto has developed Mind GPT, a model underlying a voice assistant that can understand different dialects and that can be activated without wake-up words such as "Hey, Siri." Nvidia tools are underpinning that technology.

In another Nvidia partnership, SoundHound is developing an in-vehicle voice interface that will be available even when the vehicle is offline and not connected to the cloud. Users will be able to ask questions about troubleshooting and the system will respond with answers built off the vehicle manual and other information.

AI-enabled in-car experiences will help vehicles evolve into "intelligent companions," Shapiro said. "Cars of the future will be much more than just modes of transportation."