Meta unveiled V-JEPA 2, a 1.2 billion-parameter AI world model trained on over one million hours of video which can, among other things, help autonomous vehicles understand and predict their environment—specifically the movement of physical objects. The open-source model achieved 65-80% success rates in robotic pick-and-place tasks, demonstrating capabilities crucial for self-driving cars navigating complex environments.
Subscribe to Automotive World to continue reading
Sign up now and gain unlimited access to our news, analysis, data, and research
Already a member?