Robotics' End Game: Nvidia's Jim Fan

Jim Fan, who leads the embodied autonomous research group at Nvidia, returns to AI Ascent to argue that robotics is entering its end game — and that the playbook is already written. He walks through what he calls "the great parallel": robotics following the LLM path from pre-training to reasoning to auto research, but with world models replacing language models, egocentric video replacing teleoperation, and world action models replacing the VLA paradigm. Along the way: why he thinks we'll pass the physical Turing test within 2–3 years, why "compute now equals environment equals data," and why this generation was born just in time to solve robotics. 00:00 Introduction 00:30 DGX One Origin Story 01:42 The Great Parallel 03:31 Robotics Endgame Setup 03:39 Why VLA Falls Short 04:32 Video World Models 06:09 DreamZero World Action 07:46 Scaling Data Collection 11:06 EgoScale And Scaling Laws 15:39 DreamDojo And The Roadmap

GPT-4正在为你翻译摘要中......

翻译全文