Why does it seem so obvious to us that robots should look like humans?
-
The human body has its drawbacks.
Moving through space on 2 legs is slower and less stable than walking on 4, for instance, or on 6, and the weight of the head is what causes us unnecessary back pain.
Human elbows and knees can't bend backward, and a full 360-degree shoulder rotation only happens in particularly painful accidents.
So why do we keep trying to build robots that look like us?
-
There are many answers to that.
The first is that it's what we expect to see.
Because we're used to humans, because it's what we've seen in movies, or because it feels less creepy to us. But it's what we expect to be built for us — and what we'll pay for.
The second is that the human form has its advantages too.
Moving on 2 legs allows freedom of movement in tighter spaces, a higher vantage point, and the structure of the fingers enables a range of grasping capabilities that no mechanical device can match.
But the most interesting reason has to do with software.
In order for a robot to function like a human, the underlying software is trained on a massive amount of video footage showing humans performing various actions.
This means the robot learns to behave like a human — and so the most logical step is to match its physical structure to the way it will actually behave.
In the video:
The G1 robot from Chinese company Unitree demonstrates remarkable capabilities (some of them entirely pointless). It is expected to sell for $16,000.