A few months ago, a Georgia Tech researcher unveiled a seemingly insignificant app that allows anyone to control a robot arm by tilting their phone, much like in a racing game. No training manual, no joystick, and no headset. Just a phone that people already own, aimed at a task they know how to do.
That particular detail reveals more about the future of this than any keynote address on humanoid robots could. The phone is not going away. It’s evolving into an arm-and-leg-equipped remote control.
Andy Wu, a Harvard Business School technology strategy student, puts it bluntly. He contends that most jobs are not well suited for humanoid robots, with wheels outperforming legs and specialized arms nearly always outperforming general-purpose ones. The only reason a humanoid form could prevail in any case is scale, which is also the reason your smartphone is incredibly inexpensive for what it does. The cost collapses because we produce more than a billion phones annually. Even though a wheeled robot would theoretically perform the task more effectively, the math could bend in the same way if tens of millions of humanoids were built.
It’s odd to sit with that. Superior engineering isn’t the main argument in favor of humanoid robots. It concerns whether people merely favor machines that resemble them in their homes and on factory floors. Wu refers to it as a trillion-dollar gamble based on consumer preference rather than capacity. To be honest, it’s difficult not to find that a little unsettling—we might choose the less effective robot just because it has a face.
The numbers that are floating around are striking—perhaps too so. The market for humanoid robots is expected to reach $38 billion by 2035, with unit costs expected to drop from the current $150,000 to a level that small businesses can afford. Similar trends were seen in smartphones, which went from being $500 luxury goods to being carried by almost five billion people today. It’s currently a guess disguised as a projection as to whether humanoid robots compress that timeline in the same manner.
The more difficult issues are mechanical rather than financial. Current AI allows a service robot to clear dishes and navigate a cluttered room with about 80% reliability, according to researchers at Friedrich-Alexander University in Germany. While this is useful in a care facility, it is unacceptable on an assembly line where near-perfect output is expected. Several humanoid robots completed the Beijing half-marathon last year, which sounds impressive until you consider that a robot’s legs failing after twelve miles is a maintenance catastrophe rather than a finish-line tale.
Thus, the realistic picture lies in the middle of exaggeration and disbelief. It’s unlikely that your actual phone will be replaced in the upcoming year. However, it appears to be gradually evolving into a body that can also fold laundry or replenish a shelf because of the role it plays—the object in your hand that runs errands, complies with commands, and increasingly acts on your behalf. We learned to anticipate convenience on demand thanks to smartphones. The next, clumsier, and much more costly attempt to deliver it physically is through humanoid robots.
