Musk is building them. The Chinese are building and selling them. Why?
The first answer that occurred to me is that if we end up with robots neither much abler than humans nor much less able we will want the option of employing a mixed workforce, being able to have humans and robots substitute for each other. That is easier if they are about the same size and shape. Generalizing that argument, we live in an environment created by humans for humans. Our tools are designed to fit human hands. Our rooms have space for a tall human to stand upright, shelves most of which a human of average height can reach. Our cars, buses, airplanes are designed to fit humans into. If choosing a design for a robot expected to function in that environment there is much to be said for copying the human form.
That might not be true if the robot is intended for a task for which the human form is poorly suited. The one robot I own and routinely use is my (self-driving) Tesla. Although I can imagine a very strong humanoid robot serving the same function by running with me on its back or in its arms, a car-shaped body seems an obviously better design. One could, however, replace the Tesla with a robot chauffeur driving the car I used to drive. The Tesla spends most of its time sitting in my carport; my robot chauffeur could also be my robot gardener, cook and babysitter.
The other robot that I routinely use is the AI that responds to my Google searches by telling me what it thinks I want to know, sometimes correctly. It could be replaced by a humanoid robot watching over my shoulder and responding to what I type but that would be less convenient, make less sense, than locating it invisibly online to respond to many searches by many people.
I conclude that it makes some sense for general purpose robots, robots intended for a diversity of tasks, to be humanoid. Single purpose robots such as Google’s chatbot or a self-driving car intended as a robocab, on the other hand, should have a design optimized for that task; neither a chatbot nor a robocab needs arms and legs, hands and feet.
Another advantage to a humanoid design is that most of the things we want a robot to do are things humans are already doing. We can figure out how a humanoid robot should do things, how we should train it, by free riding on information we already have about how humans do them, just as we produced chatbots from observing how humans write.[1]
A third answer is that humans have been designed by evolution as general purpose machines; if you are designing a general purpose machine it might be a good design to imitate. You would not want to copy it exactly, both because the materials and the constraints they imply are different and because the range of tasks is, but something along humanoid lines might make sense.
A fourth answer is that if we have robots we will be interacting with them and we already know how to interact with humans. A robot does not have to be shaped like a human for us to talk to it and it to talk to us but it cannot nod or shake its head if it does not have a head, smile if it does not have a mouth, shrug its shoulders if it does not have shoulders, shake hands if … . A lot of human language, broadly defined, depends on the human form.
That answer, like the first, generalizes. Our knowledge of how to interact with humans is not limited to how to communicate with them. We also know what we can expect of other humans, when we should be angry at them, when grateful, what past interactions give us reason to trust them or them us. We routinely apply that knowledge not only to people but also to things, often inappropriately, are reluctant to scrap the car that has served us well for many years because we feel it deserves better of us. Robots are more like people than are cars, may become still more like people in the future. It is easier to treat robots like people if they look like people.
My first three answers were unambiguously reasons to make robots humanoid; this one is not. If robots are not people it may be a mistake to pretend that they are. If a robot is only a machine it should be scrapped when the cost of maintenance and repairs becomes more than its services are worth, not kept going as due reward for loyal service. Shaping robots like humans encourages us to treat them as humans, sometimes a mistake.
I am now imagining a future robot rights movement. If that seems absurd, consider existing legal protections for animals and the movements that produced them. Animals are not humans either.
Of course, if AI progresses far enough, treating robots as people might no longer be a mistake.
My web page, with the full text of multiple books and articles and much else
Past posts, sorted by topic
A search bar for past posts and much of my other writing
A draft of my next book, Consequences of Climate Change, webbed for comments.
[1] China’s efforts along these lines are described in a recent Economist article.
My robot vacuum performs much better than a humanoid robot running a human vacuum.
David, nobody has mentioned what to me is the obvious answer to why we seem so committed to the idea of a humanoid robot. A large percentage of the generation that is making these decisions grew up reading Asimov’s I Robot, and watching Star Trek’s Next Generation, in which Data, the android officer on the Enterprise wished for more human traits, and was proven by Captain Picard to be a viable life form in a memorable episode during that series.
Many of the futuristic products of that tv show are now available to us. Why not a personal “Data” for every family who wants one?