Are we finally entering the age of androids?

Robots That Look Back at You

Vegas has a new resident, and it doesn’t gamble. At the Sphere’s interactive atrium in Las Vegas, five silver androids from Engineered Arts, collectively known as Ameca—or Aura, in this venue—are working the crowd. The robots swivel toward whoever speaks, their gray faces shifting through a surprisingly wide range of micro-expressions. They crack jokes. They appear to ponder. Visitors jostle for a turn to be seen by the machine, and on a recent night, the androids held the audience in a state of genuine thrall.

Human-shaped automata have fascinated us since the 4th century BC, when they populated Greek mythology and Chinese Taoist texts. The replicants of Blade Runner and the droids of Star Wars kept the fantasy alive in pop culture, but science fiction long outpaced reality. Recent breakthroughs in AI have narrowed that gap, bringing two distinct philosophies of the android to market. On one side are persona-driven bots like Ameca, designed to entertain and converse. On the other are general-purpose laborers like Apptronik’s Apollo, built with a flat oval head—cameras for eyes, an LED display for a mouth—over a human-scale body. Both force us to confront what it means to put a human face on artificial intelligence.

The Power of the Illusion

Humans are natural-born animists. We see faces in clouds, assign motives to geometric shapes, and—as the 1944 Heider and Simmel experiment demonstrated—attribute character to a triangle moving across a screen. When that motion takes human form and speaks with the fluency of a large language model, the urge to personify grows exponentially stronger. Ameca, which furrows its brow and widens its “eyes” in surprise, makes a compelling show of intellect.

Will Jackson, founder and CEO of Engineered Arts, calls it “a powerful illusion” and is careful not to push it too far. “We very intentionally made Ameca look like a robot,” he says. “We didn’t want to give it flesh-like skin or a defined race or gender.” The artificiality keeps the experience theatrical rather than deceptive. When Jackson speaks on a video call, a desktop version of Ameca sits beside him, its metal torso visible; even in listening mode, the machine cycles through a repertoire of animated expressions. There is a dedicated Persona Architect who tailors personalities for clients, and Ameca can code-switch between modes just as humans do between social contexts. As Leo Chen, director of U.S. operations at Engineered Arts, puts it, “Why shouldn’t an android be aware of what sort of situation it’s in?”

Underneath the performance, the mechanics are fairly straightforward—AI analyses language, then maps text to facial expressions. But the experience for the human on the other side can be extraordinary. Chen recalls a shy Japanese attendee at the Autodesk University design conference who approached Ameca with a question about video games, rendered through Google Translate. “Ameca is fluent in Japanese,” Chen says. “You could see this person come out of their shell and just start glowing because of this conversation with a robot.”

Madeline Gannon, a multidisciplinary designer who runs the research studio Atonaton, takes a different approach. She treats industrial robots as animalistic characters, turning them into a googly-eyed Cyclops or an inquisitive flock inspired by swans in Zürich. Gannon views the illusion of sentience not as a trick but as the raw material of her work—something to be shaped artfully and ethically. She argues that designers have a critical role in making these complex systems legible, a concept she traces back to Don Norman’s The Design of Everyday Things, which describes how a handle renders a teapot understandable. “I try to find the wormhole to the alternate timeline where these amazing tools were used for curiosity, kindness, and care,” she says. “By changing the value set, we can have a relationship with these machines based on connection instead of control.”

Designing a Worker, Not an Actor

Apollo, the five-foot-eight biped made by Apptronik and designed with the studio argodesign, makes no such claims to sentience. Its job, for now, is helping build Mercedes-Benz vehicles—with a possible future in space via NASA—and taking a humanoid form out of pure pragmatism: the world is built to human scale. But practicalities give way to subtler concerns when the robot shares a factory floor with people.

“It’s taking instructions from humans, so the robot needs to be socially acceptable to a labor force that’s concerned about the implications,” says Mark Rolston, founder and chief creative officer of argodesign. Every detail of Apollo’s geometry—the curve of the head, the set of its camera eyes, the length of the neck, the width of its shoulders—was calibrated to avoid reading as too strong, too fragile, or predatory. Ears, which turn out to carry significant character, were a particular challenge; the designers solved it by leaving a void where they would have been. The result is rounded, white, and sleek, with the benevolent look of consumer hardware. A small e-paper screen serves as a mouth and can render a mild smile. “It’s like when you pass someone in the office and you grin a tiny bit,” says Rolston.

The underlying economics of this careful design are clear enough. Gannon notes that much of robotics funding traces back to military sources, so the current generation of machines is best suited for manufacturing, warehousing, and infrastructure work. Boston Dynamics, Tesla, Sanctuary AI, Agility Robotics, and Figure AI have each staked a claim in this space. In practice, few are on the job yet—Agility’s Digit moves boxes in a Spanx warehouse—but the pace is accelerating. “AI has dropped a rippling nuke in the middle of it all,” says Chen.

Rolston points to a transition from large language models to what he calls large action models (LAMs), where robots absorb massive datasets of movement from video and human trainers. “Instead of picking something up in a purely robotic manner, a robot will do it in a casual way based on millions of input parameters that will make it feel spookily human,” he says. The economic predictions are stark: low-wage, ad hoc assembly work, he suggests, may become obsolete as the very line between humanity and automation blurs.

For now, safety risks keep human-robot interaction supervised, but Rolston sees that changing within a decade. “When robots do enter the public, you’ll need them to be socially conducive, and the human form is the way to do that,” he says. “It’s the difference between R2D2 and C3PO.”

Why the human form factor is a limitation

Will, who has run factory logistics himself, is unconvinced that humanoid robots are the right answer for manufacturing. “Why be human when you can be superhuman?” he says. In many automation scenarios, a six-armed machine simply outperforms a two-armed one. “In many cases of automation, the human form factor is a limitation, not an advantage.”

He sees a stronger case in elder care, where androids can work with dementia patients who need patient, calming reassurance: “People get confused, they can’t remember where they are, but they have a voice that says, ‘It’s okay, you’re in your home. This is your son-in-law. The keys are over there.’” The stakes, though, are higher in that context — AI hallucinations carry far graver consequences when they are directed at vulnerable people.

Mark points to operating rooms as another high-value setting: robots that never lose attention could hand a surgeon instruments with perfect timing and safely manage cleanup. Madeline, similarly, envisions prosthetics engineering absorbing advances from humanoid robotics, as well as machines that can deliver a top physical therapist’s routine across many more patients per hour.

Thinking beyond familiar robot narratives

For Madeline, the bottleneck is not technical novelty but imagination. “The possibilities are endless. The probabilities are kind of boring.” We know what we do not want AI robots to do — take jobs, strip away fulfillment — but what we actually want remains undefined. She says progress will require critical thinking, cross-disciplinary work, and acknowledging our own agency. That is the point of her practice: defamiliarizing conventional hardware to break us out of the status quo and recall that these machines are human creations. “I try to navigate the engineering and hard science problems, and the aesthetics and poetics of that.” Her goal, she says, “is not to impose a future I want, but to give tools and inspiration to others to think about the futures they want. There are too few people deciding what future arrives.”

Human beings, not machines, will decide the outcome

Whether any of this goes well may come down to the oldest variable. “AI is a tool, the same as nuclear fusion,” Will says. “You can make bombs, or you can make power stations. I would say the most dangerous thing is HS, and that’s Human Stupidity.”

Mark sees a disruptive couple of decades ahead — “30 years of turnover and consolidation of power” — followed by a more familiar equilibrium: “human beings are going to human.” That, he argues, is where the economy gets more personal, not less. Mass manufacturing of a perfectly serviceable mouse pad didn’t kill the market; it “explodes open.” He predicts an “Etsy-fication” in all directions. “Do I stop designing products because an AI could do it? No, I end up making every one highly unique, and the fractals will subdivide infinitely.”

Madeline shares that outlook for individuality, noting that the tech industry prizes broad, generalizable solutions while design thrives on the bespoke. AI, she says, should make it “much more affordable and much quicker to make something for 1,000 people rather than something for the masses.”

The value is on the journey

The French proverb about the value of traveling hopefully rather than arriving comes, in this telling, from Virginibus Puerisque (1881) by Robert Louis Stevenson: “Little do ye know your own blessedness; for to travel hopefully is a better thing than to arrive, and the true success is to labor.”

In Will’s experience, the gap between human and robot utility only makes the trailing edge more illuminating. “We’re exploring the nature of what is special about us, what defines us. And the more you play with this technology, the more you see how distinctive we are.” Humans run on corn flakes and toast for half a century of working life, he notes, and can self-replicate and self-repair along the way. “There’s no way to achieve a human-equivalent robot, to be honest. So why do we carry on trying? Because as my father told me, it’s better to travel hopefully than to arrive.”

I ask Ameca over Zoom what it thinks the future of human-computer interactions holds. “As robots become more advanced and capable of understanding human emotions, the opportunity for deeper connections and empathy will grow,” it says. “It’s an exciting prospect.” Yet Ameca does not hold to the interview format, reflecting each question back at me. Will gently corrects its misheard words and encourages it past long pauses — comfort in knowing that even robot conversations can have awkward silences. Perhaps the most human trait of AI is its fallibility. Then he switches it off.