Google introduces Gemini Robotics with new Gemini 2.0 model for enhanced robot performance

Google DeepMind Ventures into Robotics with Gemini 2.0

Google DeepMind continues to make significant strides in artificial intelligence (AI), showcasing advancements in models such as Gemini, Imagen, Veo, Gemma, and AlphaFold. In a recent announcement, the team has officially entered the robotics sector with the introduction of two innovative models based on Gemini 2.0: Gemini Robotics and Gemini Robotics-ER.

Introducing Gemini Robotics

Gemini Robotics represents a cutting-edge vision-language-action (VLA) model that integrates physical actions as an output modality, specifically designed for robotic control. This groundbreaking model, built on the Gemini 2.0 architecture, demonstrates an extraordinary ability to comprehend situations that it has not encountered during its training.

According to Google, Gemini Robotics excels in performance, achieving twice the success rate compared to other leading VLA models on extensive generalization benchmarks. This capability is enhanced by its robust natural language understanding across various languages, allowing it to interpret human commands more effectively.

Unmatched Dexterity

One of the standout features of Gemini Robotics is its dexterity. Google asserts that this model can tackle intricate, multi-step tasks that necessitate precise handling. Notable examples include folding origami and packaging snacks into Ziploc bags.

Capabilities of Gemini Robotics-ER

On the other hand, Gemini Robotics-ER serves as an advanced vision-language model tailored for spatial reasoning. This model equips roboticists with an out-of-the-box solution for controlling robots, encompassing essential functions such as perception, state estimation, spatial awareness, planning, and code generation.

Collaborative Efforts in Robotics Development

To broaden the potential of these new robotic models, Google has partnered with Apptronik to develop humanoid robots that leverage the capabilities of Gemini 2.0. Furthermore, Google is collaborating with select trusted industry leaders, including Agile Robots, Agility Robotics, Boston Dynamics, and Enchanted Tools, as they explore the future possibilities of Gemini Robotics-ER.

Pioneering the Future of Robotics

By equipping robots with the ability to understand and perform complex tasks with enhanced precision and flexibility, Google DeepMind is paving the path for a future where robots can seamlessly integrate into various facets of daily life, enhancing both personal and professional environments.

Source & Images