Are humanoid robots creepy?

Some very smart people are betting that machines shaped like humans will do much of our household and factory work for us in the near-ish future. Boosters include Elon Musk, whose eye-popping new deal with Tesla reportedly doesn’t pay out fully until the company sells a million humanoid robots. Wall Street is cheering. Goldman Sachs projects humanoid sales of $38 billion by 2035, and Morgan Stanley puts the market at $5 trillion by mid-century.

Hurdles remain. The technology is not quite baked yet, as recent headlines attest: One high-profile prototype still requires a real person to remotely operate it, another face-planted onstage at a launch event. It is proving hard to perfect the hands, which to be fair is an age-old struggle for artists as well. Let’s assume these problems get solved; as our colleague James Lambert pointed out a year ago, progress is happening at breakneck speed, and the global economic implications of getting the details right are huge.  

Yet the business case for the humanoid form factor remains in question. Machines that deftly navigate homes and workspaces without needing a redesign of those environments will be incredibly valuable. But robots don’t have to look like us or mimic all our functions to be useful, and humanoids are just one kind of embodied AI, a broad category that includes everything from autonomous vehicles to future iterations of the self-propelled vacuum cleaner now terrorizing your cat. Smart devices built for discrete tasks might prove more practical, much as single-purpose AIs challenge the primacy of expensive do-it-all models.

Another potential hangup—one of many cultural issues to be addressed as we plunge into an uncharted AI future—is that some people may be freaked out by humanoids. Everyone wants help with the laundry but not everyone is ready to live with an automaton. If you’ve ever sworn a painting’s eyes were following you across a room, you might have qualms about encountering a robot when you get up at night for a glass of water. There’s a reason R2-D2 is more popular than C-3PO. (Factories are less of a problem on this count, although next-gen facilities purpose-built for smart automation could limit demand for humanoids there.)

The uncanny valley—the name coined by roboticist Masahiro Mori for an instinctive aversion to the not-quite-human—runs deep in the collective subconscious. Two centuries and dozens of cinematic retellings after Mary Shelley wrote Frankenstein, director Guillermo del Toro’s new vision of the manmade man is a hit on Netflix. The golem of Jewish folklore, a kind of animated clay humanoid, goes back much further. A film critic for the Toronto Star wrote of the not-so-convincing computer-animated characters in 2004’s The Polar Express, “If I were a child, I’d have nightmares. Come to think of it, I did anyway.”

But sci-fi is also full of beloved humanoids, and generations to come may find a robot in the pantry no more threatening than a toaster oven. Less esoteric concerns, like unease at intelligent data-gathering units moving freely through our homes or doubts about the human warmth of a machine that tends to the elderly, could also fade in time. We’ll find out, soon.