Circuit Breakers

Inside the Canadian company betting millions on human-like robots

- MICHAEL HARRIS

THE ROBOT has a grey and immaculate face with no mouth, no nose, and two unblinking eyes. It's a face without human pretensions. But, like me, it has a name-Phoenix-and between that geometric resin mound and myself, there is an immediate camaraderie. The robot and I are basically the same height and weight; we both peer and bend toward each other with undeniable curiosity. When I extend a hand, Phoenix looks down, cocks its head three degrees, and considers. Then, gingerly, it places a hand in mine, and we shake.

"Nice to meet you," I say. "That's a good grip."

"Thanks."

The voice isn't the robot's, though; it's a human's. I turn to look at the guy strapped into a convoluted rig behind me. Eric wears an exosuit complete with a virtual reality headset and wire-spewing gloves that track his minutest movements, relaying them to Phoenix. Eric has his own hand outstretched, and he's shaking empty air. He is, in fact, remotely inhabiting the robot whose hand is still in my grip. Phoenix and Eric see together, move together.

I am only beginning to grasp what this handshake foretells. If remote-controlled greetings were the robot's whole repertoire, it would be a neat trick. But any sense of play very much misses the point. Phoenix is no puppet, no toy. And the mimicry it engages in is only a prologue to something far more profound. Phoenix is furiously-wonderfully-learning.

Built by the Vancouver company Sanctuary AI, Phoenix is racing to become the world's first general-purpose robot with human-like intelligence-a machine capable of understanding the real world and executing tasks previously done by us. It is an embodied AI. Its intelligence does not merely flash inside a server but reaches out and acts. Like a wide-eyed toddler, it gains facility as it explores and practises.