DeepMind Announces Plans to Utilize AI Models for Enhancing Physical Robots
Google’s Breakthrough in AI Robotics
On a recent Wednesday, Google DeepMind introduced two advanced AI models, Gemini Robotics and Gemini Robotics-ER, designed specifically for robotics applications. These models operate on Gemini 2.0, which Google proudly claims is its "most capable" AI system to date. The new models aim to enhance the functionality of robots, enabling them to perform physical tasks in the real world.
Collaboration with Apptronik
In a strategic partnership, Google will collaborate with Apptronik, a robotics firm based in Texas, to harness the capabilities of Gemini 2.0 for developing next-generation humanoid robots. Apptronik has previously worked with notable organizations like NVIDIA and NASA, showcasing its expertise in the robotics field. Recently, it gained attention for a funding round that raised $350 million, with Google participating in the investment.
Demonstrating New Capabilities
Google has provided intriguing demonstrations featuring Apptronik robots equipped with the new AI models. In these videos, the robots showcase their abilities by performing tasks such as plugging devices into power outlets, filling lunchboxes, and manipulating plastic vegetables. These demonstrations highlight the potential of AI-powered robotics to respond to verbal instructions and adapt to various situations in real-time.
Key Qualities of Effective AI Models
According to Google, for AI models to be effective in robotics, they must possess three essential qualities:
- General Adaptability: AI should be able to handle diverse scenarios.
- Interactivity: The model needs to respond swiftly to commands and changes in the environment.
- Dexterity: Robots must be capable of performing tasks that require human-like manipulation, such as handling small objects with precision.
By focusing on these attributes, Google aims to ensure that its AI models can contribute significantly to robotics advancements.
Gemini Robotics-ER: Foundation for Custom Solutions
Gemini Robotics-ER is tailored for roboticists looking to create custom solutions. It serves as a foundational model that can be adapted and trained for specific tasks. Google plans to make it available not only to Apptronik but also to select partners, including Agile Robots and Boston Dynamics, allowing these companies to leverage Gemini’s capabilities for their developments.
Industry Competitors in AI Robotics
Google’s strides in the AI robotics sector come at a time when other tech companies are also intensifying their efforts. Notably, OpenAI has invested in Physical Intelligence, a startup concentrated on integrating general-purpose AI into physical applications for robotics. Recently, OpenAI has made significant hires to enhance its robotics initiatives.
Moreover, Tesla has entered the humanoid robotics scene with its Optimus robot, further accelerating the technological competition in creating smart, adaptable robots.
Future Vision of AI in Robotics
Sundar Pichai, CEO of Google, shared insights on the company’s vision for the future of robotics. In his message posted on social media, he emphasized that robotics serves as an effective testing environment for translating AI technologies into tangible solutions for everyday challenges. He believes that with Google’s multimodal AI models, robots will be capable of adapting to their surroundings and making decisions on the spot, significantly improving their utility.
As advancements in AI continue to evolve, the partnership between Google and Apptronik represents a pivotal step in bridging the gap between artificial intelligence and hands-on robotics, potentially revolutionizing how we perceive and interact with machines in our daily lives.
