Google Teaches Household Robots To Keep Children And Pets Safe: 'It Can Spot Hazards'

Artificial intelligence is moving from computer screens into everyday living spaces, making it challenging for researchers to train the models so that they can coexist with humans without causing accidental harm. To tackle the same issue, Google has been training its latest robot AI to avoid harming babies and pets, according to Metro.UK.

According to the report, the DeepMind division of the company has unveiled its Gemini Robotics 2.5 model, and as per the company, it is approaching "human-level dexterity".

In a blog post on Tuesday (Nov 10), the company said that the Gemini Robotics 2.5 increases both the breadth of skills a robot can learn and the speed at which it can acquire them.

Also read | US Woman Dies After Driver Mistakes Her for Trash Bag, Drags Her 2 Miles

"As robots move into everyday environments, physical safety is key," Caroline Parada, head of robotics at DeepMind, said, as quoted by the outlet.

"With this release, the robot can spot hazards as it works and steers around them in real time, including people nearby."

She added that the researchers are aiming to make sure the robots handle everyday objects with both 'fine precision' and 'high reliability' because real-world environments are unpredictable.

She added that the researchers are expanding the model's capabilities in task breadth, performance, and complexity with Gemini Robotics 2.5.

Also read | GTA 6 Preload Causes Record Internet Traffic Ahead Of Global Launch

"We evaluated the model across 200 distinct tasks - such as serving food with tongs, tying a bow on a gift box, and pipetting a liquid," she added.

"These tasks represent a step toward human-level dexterity by combining a number of skills, including advanced tool use, complex manipulation of deformables, and handling delicate objects with finesse."

According to Google, Gemini Robotics 2.5 uses inference-time reasoning guidance to steer around physical hazards in real time, allowing the robot to navigate complex environments securely.

The blog revealed that this capability boosted safe task completion around obstacles by 33% and around nearby people by 43% in their evaluations.