The robots can learn instructions they’ve never heard before and respond with logical responses and actions.
In a demonstration last week, a researcher asked a robot, “I’m hungry, can you get me a snack?”
The robot then searched through a cafeteria, opened a drawer, found a bag of chips, and brought it to the human.
Google executives and researchers pointed out that this is the first time language models have been integrated into robots.
Brian Ichter, a research scientist at Google and one of the authors of a new paper released Tuesday describing the progress the company has made, said: "This is very fundamentally a different paradigm"
Language models work by using massive amounts of text posted online to train artificial intelligence software to predict what kinds of responses might follow specific questions or comments.
The models have become so precise at predicting the correct response that interacting with one often feels like conversing with a knowledgeable human.
The bots are already cleaning counters and throwing out trash in various Google cafeterias.
Google's robots, on the other hand, are far from ready for the real world.
Google's researchers and executives have repeatedly stated that they are simply running a research lab and have no plans to commercialize the technology.