|
Stephen Witt
A writer who covers technology—and who met more than ten robots for this story
Watching an autonomous humanoid robot come to life is one of the more stunning experiences of my career. A hand articulates, the head swivels, and the legs begin to take tentative, deliberate steps. The machine moves with circumspection, as if consciousness had just been ensouled in a new body. Driving the humanoid is the latest generation of “physical A.I.,” grown in a data center, put through millions of training runs in a virtual reality, and downloaded into a real-world robot body.
Similar scenes of digital quickening are taking place all over Silicon Valley. One memorable encounter I had while reporting this article was at 1X, a startup that is marketing its humanoid robot Neo for home delivery later this year. Neo is soft, light, and nonthreatening; in its presence, it is possible to forget that Neo is a machine. My initial impulse, upon meeting Neo, was to give it a hug.
It should be admitted that Neo’s brain has not quite caught up with its body: it fumbles simple tasks and does not always do as you ask it. That is typical of today’s humanoids, which seem to yearn for autonomy without quite achieving it. But the pace of progress is rapid, and one engineer I spoke with compared the influx of programming talent into robotics over the past year to the surge of interest in language modelling that preceded the 2022 launch of ChatGPT. Soon—maybe too soon—we will find ourselves staring into the expressionless eye cameras of our robotic doppelgängers, in search of obedience, intelligence, and perhaps even recognition.
|