Google’s DeepMind Equips Robot with Gemini for Command Execution

DeepMind’s robotics division integrated Gemini 1.5 Pro into a robot, enabling it to navigate an office and respond to commands.

In a series of videos, Google’s AI team instructs the machine to perform various tasks. In one experiment, a tester asked the robot to take him somewhere suitable for drawing. The droid led the person to a board with a marker.

View this post on Instagram

A post by Google DeepMind (@googledeepmind)

In another test, the robot was shown a map and asked to reach the “Blue Zone.” It paused briefly before proceeding with the task.

“I successfully followed the instructions on the board,” the robot commented.

For these tests, the droid was trained to navigate using “multimodal instruction navigation with demonstration tours.” It moves around the office and verbally identifies various landmarks. For instance, the robot can move to an outlet when asked to charge a phone.

The team taught the AI to reason “in accordance with common sense” based on its understanding of the environment. This enables it to respond to written and drawn commands, as well as gestures.

You can ask the droid if a favorite drink is available. Gemini knows to go to the fridge, check for cans, and return with an answer.

According to Google, the AI-powered machine demonstrated about 90% successful interactions with employees in over 50 tests.

Previously, researchers from the University of Tokyo taught a robot to drive an electric vehicle.

In July, Chinese scientists grew an artificial brain and implanted it in a droid.

Elon Musk has stated that humanoid robots Optimus will raise Tesla’s market capitalization to $25 trillion.