Speaker

Marc Raibert

Marc Raibert

Roboticist, founder and longtime CEO of Boston Dynamics, and Executive Director of the Boston Dynamics AI Institute. For over 40 years, at Carnegie Mellon, MIT’s Leg Lab, and then Boston Dynamics, Raibert has led the creation of the field’s best-known legged robots — BigDog, Spot, Atlas, and Handle — built around dynamic balance rather than the cautious, always-stable walking that earlier robotics research favoured.

Raibert converted to robotics as an MIT graduate student in 1974 after seeing a disassembled robot arm in Berthold Horn’s lab, having found neurophysiology insufficiently conceptual. His research turn toward legs began at a biomechanics conference, where he judged existing legged robots — which crept along on at least three planted feet at all times — nothing like how animals actually move. He built the field’s first hopping robots as a professor at Carnegie Mellon and MIT, founded Boston Dynamics in 1992 (initially a physics-simulation company, only later returning to robot-building), and led it through Google’s ownership and its eventual sale to Hyundai. He left day-to-day leadership of Boston Dynamics to found the Boston Dynamics AI Institute, aimed at combining the company’s mechanical and control expertise with the cognitive intelligence — planning, perception, learning from observation — he judges the field’s remaining bottleneck.

Core positions

  • Legged robots should balance dynamically, by continuously predicting and correcting motion in real time, rather than staying statically stable with several feet always planted — the design philosophy behind every robot he has built since his first pogo-stick hopper.
  • Hardware innovation still matters; treating mechanical and actuator design as a solved problem, with progress coming only from software or learning, is a mistake. His defence of hydraulic actuation — reinvented at Boston Dynamics after decades of stagnation since 1950s aviation — is a case in point.
  • Robot intelligence splits into an athletic part (movement, balance, manipulation) and a cognitive part (planning and judgement under incomplete information); athletic intelligence is largely solved, cognitive intelligence is not. See Athletic Intelligence.
  • A robotics team needs technical fearlessness, diligence, intrepidness, and technical fun — and genuine robustness is best demonstrated by showing a robot fail, not only succeed.
  • Superintelligent AI is not inherently threatening merely by being smarter than people; the relevant question is whether it remains under human control, as computers generally are today.

In the wiki