Google DeepMind gives a glimpse into a future where robots sort your trash and pack your bags

Google DeepMind is creating a more capable “brain” for robots.

Google DeepMind has already been working on ways to bring the benefits of AI into the physical world.

At the beginning of this year, the company introduced two AI models created for the robotics industry. These models allow robots to understand their surroundings and carry out physical actions.

Now, Google DeepMind has revealed the continuation of this work by presenting two new models that improve robots' “thinking.”

The first one, Gemini Robotics 1.5, “turns visual information and instructions into motor commands for a robot to perform a task,” the company explains.

This model works as the robot’s brain, assessing the information and situation before any physical action is taken. This ensures that the robot completes the tasks more clearly.

The second model, Gemini Robotics-ER 1.5, analyzes the physical world, calls digital tools, and creates steps for completing an action.

According to Google DeepMind, these models will allow developers to create “more capable and versatile robots” to carry out more difficult tasks.

Gemini Robotics-ER 1.5 is already available for developers, while Gemini Robotics 1.5 is only available for select partners.

In a dedicated video, Google DeepMind showed how these models work in reality.

In the video, the robot was asked, “Based on my location, can you sort these objects into the correct compost, recycling, and trash bins?" It was provided with a few different types of trash and three bins.

The robot was able to check the location and, according to this information, check the web for the correct recycling guidelines.

The robot then continued to sort the trash according to whether it belongs to plastic, compost, or landfill.

Both models were built on the group of Gemini models and have been provided with different datasets to make them specialize in their particular areas.

The new models are expected to help robots undertake longer tasks in different environments.

According to Google DeepMind, Gemini Robotics 1.5 has another distinctive capability that could advance the robotics industry's future.

The model can “learn across different embodiments” and transfer this knowledge from one gadget to another. This saves time and energy used to tailor a model to a new robot.

The company has also shared that the new models are being developed following AI Principles and implementing a “holistic approach to safety through high-level semantic reasoning, including thinking about safety before acting, ensuring respectful dialogue with humans.”

Unlock more exclusive Cybernews content on YouTube.