Cerence and Nvidia Debut Automotive Large Language Model CaLLM to Power New Generative AI Features
Automotive voice technology developer Cerence has introduced a new large language model (LLM) for cars. The aptly named Cerence Automotive Large Language Model (CaLLM) leverages Nvidia’s LLM technology through its Drive platform to speed up the deployment of new generative AI-powered features in both existing embedded systems and new vehicles.
CaLLM employs Cerence’s multi-billion token automotive dataset to create in-car user experiences tailored for drivers beyond the more general knowledge type of LLMs. Cerence used Nvidia’s AI foundry service to train CaLLM on its dataset. The model is highly adaptable to automotive functions, features, and requirements and can be customized for automakers through training, fine-tuning, and the creation of bespoke applications. Users can personalize the assistant and use built-in information from Cerence for its native features like Cerence Car Knowledge.
“We are delivering a breakthrough in-car computing platform that leverages our vertical tech stack to allow users to complete tasks across applications through a unified conversational interface,” Cerence CTO Iqbal Arshad said. “When paired with Nvidia’s industry-leading AI frameworks and platforms, our exclusive dataset, deep relationships with automakers, and extensive market penetration put us in the ideal position to lead with our pioneering automotive-specific LLM, CaLLM™, and our own in-car computing platform. As we look to the future, we see this proprietary architecture also benefiting industry leaders outside of automotive.”
Cerence has been a leader in car voice AI since it spun off from Nuance. The company has been keeping up with the generative AI boom as well, through projects upgrading BMW’s AI platform and embedding Microsoft Teams in its voice AI system as well. It’s also inked a deal with iHeartMedia and Radioline to incorporate their audio content into its customizable voice assistant platform and created the Cerence Exterior Vehicle Communication tool to allow drivers to control their car with their voice from outside the vehicle. CaLLM isn’t the first collaboration between Cerence and Nvidia, either. The car AI developer started supporting Nvidia’s Drive Concierge in-car infotainment system last year. CaLLM serves as the core of Cerence’s new automotive computing platform, designed to integrate new applications through conversational voice interactions.
“Generative AI and automotive LLMs are expected to enable new, intuitive, and personalized user experiences, capabilities, and features both inside and outside the car,” Nvidia vice president of automotive Ali Kani said. “By tapping Nvidia’s core expertise in cloud and edge technologies, Cerence can more quickly train, scale, and deploy these models within their in-car user-experience platform to enable safer, smarter, and more enjoyable rides.”