Alphabet has also “fired” the robots for separate collection

Teaching a robot how to open a door should bring boundless opportunities to life. However, this is not the case for one of the youngest companies controlled by alphabet, Everyday Robots. Just over a year after being out of lab X – the division of Alphabet dedicated to research projects “moonshot“, the most innovative and risky ones – the team that trained over a hundred of robots on wheels equipped with a single arm to clean the canteen tables, to separate waste collection and to open doorsis in about to close as part of the budget cuts that the parent company of Google is applying to all sectors of the company.

“Everyday Robots it will no longer be a separate project within Alphabet – said Denise Gamboa, director of marketing and communications of the company -. Some of the technology and team will be merged into existing robotics businesses within Google Research“.

Everyday Robots is theX’s last failed betwhich in the last decade has made among others aerostatic balloons capable of spreading the internet connection (Loon) and kites that generate energy (Makani), before retain them too commercially unsellable to stay afloat. Other X projects, such as Waymo (which develops self-driving vehicles) and Wing (which makes food delivery drones), are still active companies within Alphabet, even if their financial prospects are held back by regulatory and technological barriers. Like Everyday Robots, these enterprises have drawn on innovative technologies that they had shown a considerable potential in demonstrations, but not great reliability.

What was Everyday Robots

Everyday Robots has emerged from the rubble of at least eight acquisitions in the robotics sector made by Google a decade ago. The co-founders of the giant, Larry Page and Sergey Brin, they expected that machine learning would revolutionize robotics and Page in particular wanted to develop a robot dedicated to consumers, reports a Wired US a former employee of the company who spoke anonymously. In 2016, software entrepreneur Hans Peter Brøndmo was tasked with a project then known as Help (later to become Moxie for a time), which aimed to harness machine learning to develop robots that could handle routines and adapt to different environments, explains the source.

The team has set up structures where a fleet of robots repeated the same task for months, like sorting waste properly. The goal was to generate useful data to train a machine learning model that could then equip robots with the skills to use their cameras, arms, wheels and finger-like grips to interact with the world around them. The innovative aspect was that engineers were not required to stick to the traditional approach to robotics, which consists of having to program specific instructions for the machines to execute for each potential scenario. The idea proved effective for the basic tasks assigned to the machines: Google employed the fleet of Everyday Robots to clean the company’s dining rooms and check that in the peak of the pandemic the conference rooms were not cluttered.

Last year, Everyday Robots showed further progress. The company integrated a large language model similar to the one behind ChatGpt into its robotics system, allowing its mechanical helper to do tasks like bring a bag of chips to users who said they were hungry. At the time, however, Google and Everyday Robots stressed that it would still be a long time before a robotic butler could be available to consumers: changes that seem trivial to a human, such as the type of lighting in a room or the shape of the bag of chips, could in fact cause malfunctions in robots.