Google DeepMind has unveiled a major advancement in artificial intelligence for robotics with its new models, Gemini Robotics and Gemini Robotics-ER. Designed to enhance robots’ ability to adapt to complex real-world environments, these models leverage the reasoning capabilities of DeepMind’s Gemini 2.0 language model to help machines complete intricate tasks. In testing, robots trained with this AI successfully folded an origami fox, organized a desk based on verbal instructions, wrapped headphone wires, and even performed a miniature basketball slam dunk. The company has also partnered with robotics startup Apptronik to develop humanoid robots that integrate this technology.
This development marks a significant step in the race among tech giants such as Tesla, OpenAI, and Nvidia to create AI-powered robots that could revolutionize industries ranging from manufacturing to healthcare. Nvidia CEO Jensen Huang has called AI-driven robotics a multitrillion-dollar opportunity, predicting that it could become the largest technology industry in history. While robotics has traditionally struggled to transition from controlled laboratory environments to real-world applications, DeepMind’s new approach—training robots using large-scale language models—offers a promising solution. This technology enables robots to adjust to new situations, respond instantly to verbal instructions, and manipulate objects with improved dexterity, making them far more capable than previous generations of machines.
Kanishka Rao, principal software engineer at Google DeepMind, described Gemini Robotics as being twice as general as the company’s previous best models, calling it a major leap toward the goal of developing truly general-purpose robots. Experts in the field have praised the breakthrough, with MIT professor Russ Tedrake calling it an impressively comprehensive effort, demonstrating strong progress in both spatial reasoning and dexterous manipulation. Ken Goldberg, a robotics professor at UC Berkeley, noted that one of the most innovative aspects of Gemini Robotics is its ability to run smoothly in the cloud, allowing it to take advantage of Google’s vast computational resources.
Despite these advancements, experts acknowledge that general-purpose robots still have a long way to go before they become practical for widespread use. While Gemini Robotics represents a significant step forward, further refinements will be needed before AI-driven robots can seamlessly integrate into everyday life. Nonetheless, DeepMind’s latest innovation moves the industry closer to a future where intelligent robots assist humans not just in research labs but in homes, workplaces, and industries worldwide.
