Lex Fridman Podcast #97 – Sertac Karaman: Robots That Fly and Robots That Drive
May 20, 2020
Sertac Karaman, a professor at MIT and co-founder of Optimus Ride, shares his expertise in robotics and autonomous vehicles. He dives into the fascinating differences between flying and driving robots, including the potential for flying cars. The discussion touches on the crucial role of machine learning and simulation in enhancing robotic perception. Sertac also contrasts the strategies of major companies like Waymo and Tesla in the autonomous vehicle race and explores the ongoing debate between LiDAR and camera-based systems, shedding light on the future of urban transportation.
AI Snips
Chapters
Transcript
Episode notes
Flying Car Feasibility
- Utilizing underutilized agile airspace (above stone-throwing range but below large aircraft) presents significant opportunities.
- Key technical challenges include building safe, complex software and simulating human behavior.
Role of Simulation
- Simulation is crucial for developing and training robots, particularly in simulating camera input and human behavior.
- Simulating human behavior is difficult due to the lack of a clear mathematical model and reliance on potentially limited human labeling.
Human-Robot Interaction
- Autonomous vehicles are often abused because they are perceived as "things" rather than human-driven vehicles.
- This necessitates a higher level of intelligence for robots to navigate human environments effectively.