As part of X’s mission to make boring robots

These creatures are the target on the table. One of them will climb up to a table and reflect for a few seconds to see if people are sitting; if so, go ahead until you find an empty one. After stretching for a second — perhaps taking the algorithmic equivalent of a deep breath before the “Let’s Do It” moment — the robot rotates and spreads its limb, extending its arm over the table to methodically cover the surface with a clear disinfectant. He then withdraws his arm to squeeze the excess liquid into a container at its base. When the task is done, he moves on, looking for another table to make his finger.

Those who finish their meal don’t bother to look up. Robots have been doing this for weeks.

No, this is not a desperate attempt to address the labor shortage. It is the research of the author Daily Robots, X’s project, Alphabet’s so-called “moonshot factory”. The coffee testing area is home to dozens of Google campuses in Mountain View, California, where a small percentage of the company’s massive workforce has returned to work. The project hopes that the robots will be useful, operating in the wild instead of controlled environments like factories. After years of development, Everyday Robots is sending its robots out into the world, or at least from the X headquarters building, to do real work. It’s enough to invite me to observe a milestone, two years after Tom Simonite of WIRED a final look at the project. At the time, the robots were sorted in the right garbage recycling bin. Surveillance services indicate the next limit, if not the last.

I have kids, but this is a serious thing. Everyday Robots is trying to do two hard things, a challenge that is so hairy, where some question whether it’s worth the effort. The first is to perform the tasks of human helpers with credibility. Everyday Robots live on the edge of the razor Moravec-in paradox, which says that it is relatively easy for computers to do difficult cognitive work and that it is devilishly difficult to duplicate the functions of a two-year-old child. Elsewhere, under the Alphabet umbrella, robots navigate complicated traffic routes, drive cars and become safer than humans. Go champion. In the world of everyday robots, conquering a daily task, such as crossing a confusing room and opening a delicate door handle, is like winning the Super Bowl. The activity of deleting tables, for example, is not just a finger; includes the full set of actions. Take into account what happens when a human or object blocks the path. “It’s the right answer for the robot, well, am I good enough to move around in an elegant place?” says Darcy Grinolds, who heads the team to validate the project’s hardware reliability and design. “Or should I completely redirect myself?”

The second hard thing the project is trying to do is move towards that goal so that it makes more sense, economically and efficiently, than a bored and paid human being in the hands of a robot.

Google, and now X, has been obsessively pursuing this approach for more than a decade. The head of the Everyday Robots team is Norwegian engineer Hans Peter Brondmo, a businessman and engineer who joined X in 2015, and had to make sense of a cacophony of robotic purchases by former leader Andy Rubin, who left the company. a cloud of sexual harassment claims. “Hans Peter wasn’t a clear choice,” says Astro Teller X CEO. “He cares about robotics, but he would be the first person to tell you that he’s not a global robot. I chose him because he’s a global entrepreneur who really understands people. And he’s a wool-dyed socialist, he’s from Norway! ‘

In an office he shares with a non-functional robot arm built in adolescence, Brondmo explained that making an effective general-purpose robot was only possible with the latest advances in machine learning. Engineers use machine learning to train their knowledge of software objects and then run millions of simulations to compress test hours into weeks. This helps their woody lab robots truly understand their environment, and they build that knowledge by accumulating a set of tools that help them solve the dilemmas they need to deal with in the wild. Although everyday robots aren’t as bright as dystopian androids in Boston Dynamics videos, they’re optimized for things to do. (Alphabet once had Boston Dynamics, but he sold in 2017.)