Lex Fridman PodcastSertac Karaman: Robots That Fly and Robots That Drive | Lex Fridman Podcast #97
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
At a glance
WHAT IT’S REALLY ABOUT
Sertac Karaman maps our robotic future: cars, drones, society, tradeoffs
- Lex Fridman and MIT professor/Optimus Ride co‑founder Sertac Karaman discuss the technical and societal challenges of deploying robots that both drive and fly. They contrast autonomous cars versus drones at scale, exploring perception, simulation, human behavior modeling, and the role of machine learning. Karaman explains Optimus Ride’s geo‑fenced, human‑in‑the‑loop autonomy strategy and contrasts it with Tesla’s and Waymo’s approaches, including safety, business models, and data. The conversation closes on high‑speed autonomous drone racing, hardware–software co‑design, and Bellman’s equation as a foundational yet computationally daunting idea in decision‑making.
IDEAS WORTH REMEMBERING
5 ideasScaling autonomy in human environments is harder than isolated robotics.
Industrial robots and Mars rovers work in structured or remote settings; putting thousands of autonomous cars or drones into everyday human spaces adds layers of algorithmic, legal, business, and social complexity that we haven’t solved at large scale.
Autonomous cars will likely reach dense deployment before autonomous air transport.
Ground vehicles can leverage and modestly adapt existing infrastructure and safety paradigms, whereas filling the ‘agile airspace’ with passenger drones involves harder safety, airspace management, and technology certification problems.
Simulation’s biggest frontier is realistic humans, not just physics and sensors.
We can now simulate dynamics and cameras increasingly well, but convincingly simulating human appearance and behavior—and using that to train and test autonomy—remains a core unsolved bottleneck.
Human behavior prediction and game‑theoretic interaction are central to safe autonomy.
Knowing where others are is no longer enough; vehicles must predict what humans will do and understand social cues (aggression, deference, signaling) while also handling being ‘abused’ as non‑human agents in human spaces.
Human‑in‑the‑loop fleets can bridge the gap to full autonomy.
Optimus Ride pushes a model where a small control center staff supervises many vehicles, intervening for efficiency rather than safety; this enables higher speeds and better service now, without waiting for perfect end‑to‑end autonomy.
WORDS WORTH SAVING
5 quotesThe real challenge of our time is to take these vehicles and put them into places where humans are present.
— Sertac Karaman
We may almost see a whole mushrooming of this technology in all kinds of places that we didn’t expect before, and that may be the real surprise.
— Sertac Karaman
You either put money for the lidar or you pay money for the compute. If you don’t put the lidar, it’s a more expensive system because we have to put in a lot of compute.
— Sertac Karaman
I’d love to build autonomous vehicles like drones that can go far faster than any human possibly can.
— Sertac Karaman
There are some things that are so far ahead people think they’re close, and there are things that are actually close people think are far ahead.
— Sertac Karaman
High quality AI-generated summary created from speaker-labeled transcript.
