Why autonomous vehicles should learn to talk to themselves

Why autonomous vehicles should learn to talk to themselves Autonomous vehicle companies are trying a new approach to training artificial intelligence to navigate roadways — getting vehicles to talk to themselves. As companies push for wider deployment of AVs, they are employing large language models, a deep learning tool that can predict and generate language, to create a vehicle's inner monologue that it can use to reason through novel situations. "If we think about what's required to build the Level 5 dream, it's going to have to be a system that has that general purpose level of reasoning, and I think there is an opportunity to learn that capability from language," said Alex Kendall, CEO of AV company Wayve, at Nvidia's GTC conference last week. Level 5 vehicles are those that can drive themselves without human supervision. Typically, programmers have used large language models to create more lifelike and responsive voice assistants and customer service bots in the auto industry. In some cases, they are being used for live driving narration, so a vehicle can explain its actions to passengers. But the next frontier is essentially teaching vehicles to use language to parse unfamiliar situations — similar to how many humans do. — Molly Boigon Read the full story here. What you need to know

Why autonomous vehicles should learn to talk to themselves

Featured Stories