Lex Fridman PodcastMarc Raibert: Boston Dynamics and the Future of Robotics | Lex Fridman Podcast #412
At a glance
WHAT IT’S REALLY ABOUT
Marc Raibert on athletic robot intelligence, hardware, and daring design
- Marc Raibert traces his 40+ year journey from early hopping robots and MIT’s Leg Lab to founding Boston Dynamics and now leading the Boston Dynamics AI Institute. He explains why dynamic, aggressive movement and continual hardware innovation are central to lifelike robots, contrasting this with the cautious, quasi-static approach common in robotics. The conversation dives into BigDog, LS3, Spot, Atlas, and dancing robots, emphasizing robustness through brutal real-world testing and iterative failure. Raibert also outlines his new focus on combining “athletic” and “cognitive” intelligence so robots can watch humans, understand tasks, and then perform them autonomously.
IDEAS WORTH REMEMBERING
5 ideasDynamic, ‘athletic’ movement is essential for lifelike robots.
Raibert argues that animals and humans move by balancing, flying phases, and reusing energy like springs—not by creeping with multiple legs always on the ground—so robots must embrace dynamic, underactuated motion to approach natural performance.
Hardware innovation still profoundly matters in robotics.
Contrary to views that software and AI alone drive progress, he stresses that advanced actuators, lightweight structures, integrated hydraulic power units, and clever mechanical design (e.g., valves in Atlas, passive dynamics) enable the control and behaviors we see.
Real-world robustness comes from aggressive, repeated failure.
Boston Dynamics’ motto of “build it, break it, fix it” led to robots enduring hundreds of falls and harsh tests (like tugging ropes, pushing doors) to expand their operational envelope; reliability is engineered through systematically exploring and surviving edge cases.
Athletic and cognitive intelligence must be combined for useful robots.
Today’s robots are physically capable but “pretty dumb,” requiring experts to script behaviors; the AI Institute aims for systems that can watch a human perform a task, segment and understand it, and then execute it—“watch, understand, do”—across varied environments.
Good teams depend on four cultural pillars: fearlessness, diligence, intrepidness, and fun.
Raibert looks for engineers who take on unsolved problems, insist on broad, robust solutions, persist through long stretches of failure, and genuinely enjoy technical work, creating an environment where ambitious robotics is sustainable.
WORDS WORTH SAVING
5 quotes“People who think you don't need to innovate hardware anymore are wrong.”
— Marc Raibert
“Most robots are pretty dumb… If robots are gonna satisfy our dreams, they need to be smarter.”
— Marc Raibert
“We had a motto at Boston Dynamics: you have to run before you can walk.”
— Marc Raibert
“Technical fearlessness means being willing to take on a problem that you don't know how to solve.”
— Marc Raibert
“Engineers get to make stuff that didn’t exist before… it’s really a higher calling.”
— Marc Raibert
High quality AI-generated summary created from speaker-labeled transcript.
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome