Marc Raibert: Boston Dynamics and the Future of Robotics | Lex Fridman Podcast #412

Marc Raibert traces Boston Dynamics from a $250K DARPA pogo-stick grant to Atlas backflips, and argues hardware innovation is still the bottleneck in robotics.

• BigDog was designed to carry 400 lb but was tested carrying ~1,000 lb, including one BigDog carrying another.
• Boston Dynamics was bootstrapped for 20 years with no outside investors until Google acquired it.
• DARPA funded the first legged robot research with $250,000 in 1980 after a cold-office pitch by Raibert and Ivan Sutherland.
• Atlas took 109 tries to nail its parkour sequence; the team has video of every attempt and it took six weeks.
• Raibert sees robot manipulation as 50 years behind legged locomotion because the field is too focused on static, safe grasping rather than dynamic handling.
• Warehouse automation is currently the only robotics use case generating real money; social/home robots have seen multiple company failures.
• Raibert is not worried about AGI risk, comparing it to Oppenheimer’s chain-reaction fears — real but not worth halting progress over.

Guests: Marc Raibert, founder and former CEO of Boston Dynamics; Executive Director, Boston Dynamics AI Institute · 2024-02-16 · Watch on YouTube