DeepMind Announces Plans to Utilize AI Models for Enhancing Physical Robots
Google DeepMind Launches New AI Models for Robotics
Introduction to Gemini 2.0
Recently, Google DeepMind introduced two cutting-edge AI models tailored for robotics: Gemini Robotics and Gemini Robotics-ER (Extended Reasoning). Operating on the Gemini 2.0 platform, these models represent what Google claims to be its most advanced artificial intelligence to date. Unlike traditional generative AI, which excels at text and image outputs, Gemini Robotics focuses on enabling robots to perform physical tasks through actionable commands.
Collaboration with Apptronik
To further its goal of enhancing robotics, Google is collaborating with Apptronik, a robotics firm based in Texas. This partnership aims to develop the next generation of humanoid robots using Gemini 2.0 technology. Apptronik has a history of collaboration with notable organizations, including Nvidia and NASA, showcasing its expertise in robotics development. Recently, Google participated in a $350 million funding round to help fuel Apptronik’s growth.
Demonstration of Capabilities
In a series of demonstration videos, Google showcased the potential of Apptronik’s robots powered by the new AI models. These robots performed several tasks, including:
- Plugging items into power strips
- Filling a lunchbox
- Manipulating plastic vegetables
- Zipping up bags
The robots responded effectively to voice commands, showcasing a significant leap in how robots might assist people in everyday tasks. However, Google did not specify when we can expect these innovations to become available to the public.
Key Qualities of Effective Robotics AI
For Google, the success of robotics powered by AI hinges on three essential characteristics:
- Generalization: Robots must adapt seamlessly to a variety of situations.
- Interactivity: They should quickly understand and respond to instructions and environmental changes.
- Dexterity: AI should enable robots to perform tasks similar to how humans manipulate objects with their hands and fingers.
Gemini Robotics-ER for Development
The Gemini Robotics-ER model is specifically designed as a foundational tool for roboticists to create and train their own AI models. This model is available not only to Apptronik but also to a select group of "trusted testers," including companies like Boston Dynamics, Agile Robots, and Agility Robotics.
The Expanding Landscape of AI in Robotics
Google is not the only tech giant pursuing advancements in robotics through AI. OpenAI has recently invested in Physical Intelligence, a startup dedicated to merging general-purpose AI with robotics to enhance their efficiency and applicability. Furthermore, OpenAI is expanding its robotics team by recruiting leaders from successful initiatives, such as Meta’s augmented reality team.
In addition, Tesla is making strides in humanoid robotics with its Optimus robot. This suggests a growing competition among major companies to integrate AI into robotics, pushing the boundaries of what these machines can achieve.
Google’s Vision for AI in Robotics
Sundar Pichai, the CEO of Google, emphasized the significance of robotics as a platform to test and refine AI advancements for real-world applications. According to Pichai, the robots powered by Google’s multimodal AI models will be capable of making adjustments in real-time and adapting to their surroundings, thereby enhancing their utility in various contexts.
With these developments, Google is paving the way for a future where AI and robotics work in tandem, potentially transforming our daily lives.
