DeepMind Announces Plans to Utilize AI Models for Physical Robots
Google Unveils Advanced AI Models for Robotics
Introduction to Gemini 2.0
On Wednesday, Google DeepMind introduced two innovative AI models specifically designed for robotics, both operating under what’s known as Gemini 2.0, which Google considers its most advanced AI framework to date. This breakthrough aims to merge artificial intelligence with physical robotics, moving beyond text and image generation to enabling robots to perform tangible tasks.
Partnership with Apptronik
In an exciting development, Google has partnered with Apptronik, a robotics company based in Texas, to develop next-generation humanoid robots powered by Gemini 2.0. Apptronik has a solid track record, having collaborated with notable organizations like Nvidia and NASA. Recently, Google also participated in a $350 million funding round for Apptronik, showcasing the tech giant’s commitment to advancing the field of robotics.
Demonstrations of New Capabilities
Google released demonstration videos displaying Apptronik robots equipped with the latest AI models, showcasing their ability to perform various tasks based on verbal commands. Some of the actions highlighted included:
- Plugging items into power outlets
- Filling a lunchbox with objects
- Moving plastic vegetables
- Zipping up bags
These demonstrations illustrate the versatility and potential usefulness of these AI advancements in everyday scenarios, although Google has not shared a timeline for when these technologies will be available for public use.
Essential Qualities of Robotics AI
According to Google, for AI models to effectively support robotics, they need to exhibit three primary characteristics:
- Generality: The ability to adapt to diverse situations and tasks.
- Interactivity: Quickly understanding and responding to spoken instructions or environmental changes.
- Dexterity: Performing actions akin to human hand movements, enabling manipulation of various objects with precision.
Gemini Robotics-ER for Developers
The Gemini Robotics-ER model is intended for roboticists seeking a strong foundation to train their own AI models. This advanced tool will be accessible not only to Apptronik but also to a select group of "trusted testers," which includes leading companies in the robotics sector, such as Agile Robots, Agility Robotics, Boston Dynamics, and Enchanted Tools.
Competitors in Robotics AI
Google’s entry into robotics AI is part of a broader industry trend. In late 2023, OpenAI made headlines by investing in Physical Intelligence, a startup centered on integrating general-purpose AI into robots. OpenAI’s engagement involved hiring a key figure from Meta’s augmented reality division to lead efforts in developing robotics and consumer hardware.
Tesla is another formidable player in the humanoid robotics arena, actively developing its own robot called Optimus. This growing competition emphasizes the increasing focus on utilizing AI in robotics.
Vision for Robotics
Google’s CEO Sundar Pichai highlighted the strategic importance of robotics as a testing ground for applying AI developments in real-world scenarios. He mentioned that these robots will leverage Google’s multimodal AI models to adjust and adapt instantly to their surroundings. This adaptability may pave the way for the next phase of intelligent machines that can seamlessly interact with their environments.
As robotics technology continues to evolve, the ways in which AI can enhance physical tasks present exciting possibilities for the future of various industries and everyday life. With companies like Google, OpenAI, and Tesla leading the charge, the landscape of robotics is set for a significant transformation.
