Gemini Robotics – A Step Towards Human-Aided Machine Learning for Human-Assisted Robotic Control and Robotic Propagation

The model known as Gemini Robotics — announced on 12 March in a blog post and technical paper — is “a small but tangible step” towards that goal, says Alexander Khazatsky, an AI researcher and co-founder of CollectedAI in Berkeley, California, which is focused on creating data sets to develop AI-powered robots.

Parada explains that when you have items on a table in front of you, you need to know how to open them, where to place them, as well as where to put them. That’s the kind of reasoning Gemini Robotics-ER is expected to do. The system that controls a robot’s movements is the system that roboticists can connect with to enable new capabilities.

We have made progress with each of these areas individually in the past but we are going to dramatically increasing performance with a single model. “This enables us to build robots that are more capable, that are more responsive and that are more robust to changes in their environment.”

In the last few years, it has become apparent that there is hope for a revolution in Robotics but there are still a lot of obstacles to overcome.

Artificial intelligence can often be used to power all sorts of clever, capable, and occasionally homicidal robots. An important limitation of the best artificial intelligence today is that it is locked 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.

A specialized version of the model was created to help in reasoning tasks with 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.