Published on: March 20, 2025 / update from: March 20, 2025 - Author: Konrad Wolfenstein
Google Gemini 2.0, The artificial intelligence and robotics: Gemini Robotics and Gemini Robotics-Er-creative image: Xpert.digital
Deepmind presents Gemini: The next era of robotics begins
Gemini Robotics: Google's transformative merger of artificial intelligence and robotics
On March 12, 2025, Google Deepmind presented its latest project Gemini Robotics, an impressive technology that combines the powerful Gemini 2.0 language model with advanced robotics. This innovation marks an important milestone in the development of intelligent robot systems that can understand natural language and perform complex physical tasks.
Google Deepmind is a leading research company for artificial intelligence (AI), which was founded in 2010 and taken over by Google in 2014. It focuses on the development of advanced AI technologies, which are characterized by neural networks with short-term storage and artificial memory. Deepmind has achieved significant breakthroughs, including defending human players in the game “Go” and the development of Alphafold, a system for predicting protein structures. Deepmind's technologies are used in areas such as robotics, medicine, energy efficiency and language processing.
The technological foundations of Gemini Robotics
Gemini Robotics was designed as a progressive vision length of the model (VLA) model, which builds on the already powerful Gemini 2.0. The central innovation is that the system can not only process digital data such as texts, images or videos, but can also perform physical actions in the real world for the first time.
The technology uses the multimodal understanding of Gemini 2.0 and expands it with a decisive new modality: physical actions. This enables the robots to bridge the digital and physical world in a way that was not yet possible.
Suitable for:
Functionality and perception skills
The technological breakthrough of Gemini Robotics lies in its ability to perceive the surroundings through cameras, to recognize objects and to capture their spatial dimensions. This information is then converted into a 3D world with precise technical coordinates.
The system can also:
- Understand natural language commands and implement it in physical actions
- Understand complex spatial relationships between objects
- Adapt to new, unknown situations
- Generate over different robot types
The two complementary models: Gemini Robotics and Gemini Robotics-Er
Google Deepmind has not only presented one, but two specialized models that address different aspects of the robotics AI.
Gemini Robotics
The main model Gemini Robotics combines Gemini 2.0's language processing skills with physical control. It enables robots to react to naturally language commands, understand complex environments and carry out adaptive actions.
Gemini Robotics
The second model, Gemini Robotics-Er (whereby he stands for “Embodied Reasoning” or “Modified Logic”), focuses on improved spatial thinking. This ability is crucial for robots that have to act in dynamic, three -dimensional environments.
Gemini Robotics-Er, for example, can intuitively recognize how an object can best be used. If a coffee cup is shown to the model, it can independently choose a suitable two-finger handle to lift the cup on the handle and calculate a secure movement.
Demonstrated skills and practical applications
In impressive demonstration videos, Google Deepmind shows the practical skills of the new AI models. The robot systems can carry out a variety of complex tasks, including:
- Folds of origami and paper
- Sorting and organizing objects based on verbal instructions
- Precise gripping and moving fragile objects
- Careful insertion of glasses in an etui
- Dice and manipulate small objects
- Closing a zipper together
- Wrapping headphone cables
- Execution of precision tasks such as basketball dunking
It is particularly noteworthy that the robots perform these tasks autonomously after they have only received an instruction. The system independently detects objects, identifies them, derives the necessary individual steps and controls the robot arms accordingly.
Strategic partnerships for further development
In order to open up the full potential of this technology, Google Deepmind works with leading companies from the robotics industry:
- Apptronik, a Texan start-up that has developed the humanoid robot “Apollo”, which is designed for logistics and manufacturing tasks such as lifting, moving and stacking of boxes
- Boston Dynamics, a well-known robotics company that was Ironically bought by Google and sold again later
- Agility Robotics and Agile Robots as other partners for the development and test of Gemini Robotics-Er
This cooperation shows Google's strategy to implement and test the technology on various robot platforms to ensure their broad applicability.
Suitable for:
Meaning for the future of robotics
The director of robotics at Deepmind, Kanishka Rao, said during a press conference, one of the greatest challenges in robotics consist that robots typically work well in known scenarios, but fail in unknown situations. Gemini Robotics should solve exactly this problem.
Suitable for:
The integration of Large Language Models (LLMS) into the robotic is part of a growing trend, and Gemini's approach could be one of the most impressive examples of this. Jan Liphardt, professor of bio engineering at Stanford University and founder of OpenMind, emphasizes that this is “one of the first examples of the use of generative AI and large language models on advanced robots” and “really the key to the development of robot helpers and robot companions” could be.
Nvidia CEO Jensen Huang goes even further and indicates that the use of generative AI to provide robots could be a market potential of several trillion US dollars on a large scale.
Gemini and robotics: A turning point for intelligent systems?
Despite the impressive progress, there are still challenges. Ken Goldberg, professor of robotics at the University of California in Berkeley, describes the AI systems as “an exciting development in the field of robotics”, but points out that “there is still a lot to do before all-purpose robots are ready for use in everyday life”.
Google plans to give further insights into the possibilities of this technology around the upcoming Google I/O conference. With his many years of interest in robotics and now with Gemini as a suitable software component, Google could open a new chapter in the development of intelligent robots.
From language to action: Google sets new standards in robotics
With Gemini Robotics, Google Deepmind has taken an important step towards the fusion of AI and robotics. The ability to understand natural language, perceive complex environments and carry out physical actions could revolutionize the way in which robots will be used in the future.
This technology marks the transition from purely digital AI applications to systems that can have a direct impact on the physical world. While this may trigger concerns with some AI skeptics, Google Deepmind's main focus is on developing adaptive and useful robot systems that can manage complex tasks with less training.
The coming years will show how this technology is developing and what practical applications you will find in different areas, from industry to everyday life.
Suitable for:
Your global marketing and business development partner
☑️ Our business language is English or German
☑️ NEW: Correspondence in your national language!
Konrad Wolfenstein
I would be happy to serve you and my team as a personal advisor.
You can contact me by filling out the contact form or simply call me on +49 89 89 674 804 (Munich) . My email address is: wolfenstein ∂ xpert.digital
I'm looking forward to our joint project.
