Your dishwasher does not look like a human. It has no arms, no hands, no face. It does not pick up the plates one by one to scrub them with a sponge. And yet it washes the dishes better and faster than you do.
So why is it that, when we imagine the robots of tomorrow, we almost always imagine them with two arms, two legs and a head?
The practical argument
Sam Altman has a simple answer, which he gave when announcing that OpenAI would build a humanoid robot: the world is made for humans. Doors, stairs, tools, cars. A robot that wants to get around everywhere and use everything is better off having our shape.
The argument is sound. But it does not explain everything. The most useful robots today, those in factories, warehouses or kitchens, almost never have a human form. When you design a machine for a specific task, you give it the shape of the task, not ours.
A very old fascination
There is probably something else. The urge to make beings in our own image is one of the oldest in humanity.
Legends are full of it, from the Golem shaped from clay to statues that come to life. In the 18th century, the French inventor Jacques de Vaucanson fascinated Europe with his automata, including a flute player capable of playing real tunes. Long before electricity, we were already dreaming of machines that resemble us.
The word "robot" was born in 1920, in a play by the Czech writer Karel Čapek, R.U.R. It comes from the Czech "robota", which means forced labour, drudgery. In the play, artificial beings made to work in place of humans eventually revolt. From its birth, then, the robot carried two contradictory dreams: that of a servant who resembles us, and the fear that, from resembling us so closely, it might one day want the same things we do.
What resembles us moves us
There is an even more troubling reason. We react to what resembles us. A face, a gaze, a voice: our brain is built to grow attached to them, almost in spite of ourselves.
In the 1960s, the computer scientist Joseph Weizenbaum created ELIZA, a very simple program that imitates a psychotherapist by rephrasing its interlocutor's sentences. He recounted being struck to see his own secretary, who knew how the program worked, ask him to leave the room so she could talk to it one to one.
A human-shaped robot will provoke these reactions with far greater force. That is not necessarily a bad thing. But it is a power, and those who make these machines know it. We discussed this on Saturday in relation to virtual companions we grow attached to.
The uncanny valley
This tendency does have a limit, however. In 1970, the Japanese roboticist Masahiro Mori observed that the more a robot resembles us, the more we like it... up to a point. When it becomes almost human, but not quite, it suddenly starts to make us uneasy. He called this dip the "uncanny valley".
We want machines that resemble us, but not too much. Enough to reassure us, not enough to unsettle us.
The takeaway
The human form of robots meets a practical need. But it also meets a deeper, older desire: that of not being alone with our machines.
A robot that resembles us forces us to ask a question the dishwasher never asks: do we expect a tool from it, or a fellow being?