Inside X’s mission to annoy the robot


These animals target tabletops. One of them would wheel up to a table and think for a few seconds about whether people were sitting; If so, move on until it is found empty. After lingering for a second – perhaps taking a deep breath algorithmic equivalent before the “let’s do it” moment – the robot rotates and exposes its limbs, extending its arms over the table to systematically cover the surface with a clean disinfectant. It then withdraws the hand to draw out the excess liquid in a bucket at its base. Task completed, it moves forward, looking for another table to swipe.

People don’t even bother to look after their lunch. The robots have been doing this for weeks.

Everyday Robot X has built more than 100 robots at Mountain View headquarters.

Photo: Michelle Groscope

No, it is not a desperate attempt to meet the labor shortage. This is a project of Alphabet’s self-styled “Munshot Factory” X, researched by everyday robots. Caf Testing Ground is one of the few dozen on Google Campus in Mountain View, California, where a small percentage of the company’s huge workforce is now back in business. The project hopes to make the robot fit by working in a wild environment instead of a controlled environment like a factory. After years of evolution, everyday robots are finally sending their robots into the world — or at least outside the X headquarters building করার to do real work. It’s a milestone enough that they invited me to observe, two years after WIRED’s Tom Simonite last saw the project. At the time, they had robots sorting garbage in the right recycling bin. The doorman services represent the next, if not the final, border.

Squeeze Robot Attack: X’s Fleet wipes the table at a Google Cafe.Video: Wired stuff

Darcy Greenlands leads the daily hardware reliability and design verification team.

Photo: Michelle Groscope

I’m a kid, but this is serious. Every day two robots are trying to work really hard, a challenge so hairy that some people question whether the effort is worth it. The first is faithfully performing the work of human helpers. Everyday robots live on the edge of the razor in Moravic’s paradox, which states that it is relatively easy for a computer to perform difficult cognitive tasks and that it is demonically difficult to mimic the functions of a two-year-old. Elsewhere, under the alphabet umbrella, robots navigate complex traffic routes, drive automobiles safer than humans, and become champions of Go. In the world of robotic robots, conquering a mundane task, such as crossing a chaotic house and opening a hard door handle, is like winning the Super Bowl. For example, the table wiping activity is not just a swipe-it includes a complete suite of actions leading to it. Take what happens when the path is blocked by a person or object. “The right response for the robot is, okay, do I have enough space to get around nicely?” Darcy Greenholds, who is leading the project’s hardware reliability and design verification team, said. “Or do I have to reorganize myself completely?”

The second difficult task the project is trying to do is to move towards a goal that makes it more understandable to have a robot in hand than a boring and low-paying person, both economically and efficiently.

A robot opens the door to the future.

Video: Wired stuff

Google, and now X, has been pursuing this vision for over a decade. The daily robot team is led by Hans Peter Brandmo, an engineer of Norwegian descent, an entrepreneur and engineer who joined X in 2015 and had to deal with a fraudulent acquisition of robotics by former leader Andy Rubin, who left the company in a sex cloud. “Hans Peter was not an obvious choice,” said Astro Taylor, CEO of X. “He thinks about robotics, but he’s the first person to tell you that he’s not a world-class robotist. I chose him because he is a world class entrepreneur who really understands people. And he’s a kind of colored socialist – he’s from Norway! “

In an office he shares with a dysfunctional robot arm he built as a teenager, Brandmo explains that the creation of an effective general-purpose robot has been possible only through recent advances in machine learning. Engineers train the software to recognize objects using machine learning and then run millions of simulations to compress a few weeks of testing into hours. This helps his laboratory robots to truly understand their environment and collects a toolset based on that knowledge that helps solve the inevitable dilemmas of dealing with wildlife. While the everyday robots may not be as glamorous as the dystopian androids in Boston Dynamics videos, they are optimized for getting things done. (Alphabet was once owned by Boston Dynamics, but sold out in 2017.)



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *