A computer model has made its way into the physical world

DeepMind: Exploring the Brain for Robotic Robots via a Large-Scale Language Model with Artificial Intelligence and Neural Networks

An artificial-intelligence company called DeepMind has made a version of its most advanced large language model into robots. Using the model, machines can perform some tasks — such as ‘slam dunking’ a miniature basketball through a desktop hoop — despite never having watched another robot do the action, says the firm.

On tasks that came up during training and unfamiliar activities, the team tested the humanoid-robots and robotic arms of Gemini Robotics. According to the team, robots using the model consistently outperformed state-of-the-art rivals when tested on new tasks and familiar ones in which details had been changed.

“While we have made progress in each one of these areas individually in the past with general robotics, we’re bringing [drastically] increasing performance in all three areas with a single model,” Parada said. This helps us to build more capable, responsive and robust robots that are more robust in their environment.

In the past few years, huge leaps in technology have led to hopes of aRobot revolution, but still big hurdles remain.

Artificial intelligence is capable of lots of different things in sci-fi tales. The best artificial intelligence is trapped inside the chat window.

A team at Google DeepMind, which is headquartered in London, started with Gemini 2.0, the firm’s most advanced vision and language model, trained by analysing patterns in huge volumes of data.

They designed a model that would excel at reasoning tasks using 3D physical and spatial understanding.

Finally, they further trained the model on data from thousands of hours of real, remote-operated robot demonstrations. This allowed the robotic ‘brain’ to implement real actions, much in the way LLMs use their learned associations to generate the next word in a sentence.