Will We Ever Trust Robots?

Read more: See here

The world might seem to be on the brink of a humanoid-robot heyday. New breakthroughs in artificial intelligence promise the type of capable, general-purpose robots previously seen only in science fiction—robots that can do things like assemble cars, care for patients, or tidy our homes, all without being given specialized instructions. 

Yet recent progress has arguably been more about style than substance. Advancements in AI have undoubtedly made robots easier to train, but they have yet to enable them to truly sense their surroundings, "think" of what to do next, and carry out those decisions in the way some viral videos might imply. In many of these demonstrations (including Tesla's), when a robot is pouring a drink or wiping down a counter, it is not acting autonomously, even if it appears to be. Instead, it is being controlled remotely by human operators, a technique roboticists refer to as teleoperation. The futuristic looks of such humanoids, which usually borrow from dystopian Hollywood sci-fi tropes like screens for faces, sharp eyes, and towering, metallic forms, suggest the robots are more capable than they often are.

"I'm worried that we're at peak hype," says Leila Takayama, a robotics expert and vice president of design and human­-robot interaction at the warehouse robotics company Robust AI. "There's a bit of an arms war—or humanoids war—between all the big tech companies to flex and show that they can do more and they can do better." As a result, she says, any roboticist not working on a humanoid has to answer to investors as to why. "We have to talk about them now, and we didn't have to a year ago," Takayama told me.

This is about more than just Alfie's appearance. Hashme and his colleagues are envisioning the way the robot moves and signals what he'll do next; imagining desires and flaws that shape his approach to tasks; and crafting an internal code of ethics that governs the instructions he will and will not accept from his owners.