Firstly, it shows we’re close to breaking another better-than-nature barrier when it comes to robotics and AI. This time in the guise of a metallic quadruped that sports a running efficiency similar to that of its furry analogue. That’s significant because we’ve already had machines best flesh in many cognitive tasks, yet the seemingly mundane activity of walking around has taken many decades longer to crack.
Which brings me to the second reason: we only accomplished this feat by mimicking nature. Certainly, there might be yet undiscovered configurations that are even more efficient, but this quadruped design and decidedly feline gait have helped the MIT researchers to achieve their efficiency goal. Nothing can really beat a few hundred million years of clumsy trial and error by evolution in refining a design to staggering (or galloping) efficiency. That, and some very clever high-torque-density electric motors of more recent origin.
Thirdly, this, in turn, reminds us that raw computation isn’t all there is to intelligence, and the embodied stuff that we take for granted is in many ways far more interesting. The tendons, the angles of the legs, the way they snap forward after taking a step – all these things are built into the physical leg itself rather than being governed top-down from some centralised information processor. Embodied cognition on the trot.
See the robot cheetah here: