Automated Assistance Arrives on Robots Through Google's Gemini Project
Step into the tech-driven future, my friend! I've barely managed to get Gemini running on my Android, and Google's already unveiling Gemini 2.0 for real-life robots. They've revealed two new AI models that Google claims will lead the charge for a new generation of helpful robots - and these robots look like us!
Enter Gemini Robotics - an advanced vision-language-action (VLA) model based on Gemini 2.0, the very same one I've been feeding PDFs to and grilling about horoscopes. This upgraded version introduces physical actions as the outcome of a query, which on my Pixel phone results in an action or answer. In a robot, however, Gemini interprets these commands as something it must physically respond to.
Google's second AI model is Gemini Robots-ER, a vision-language (VLM) model with enhanced spatial understanding, also known as "embodied reasoning." This feature allows the AI to navigate its environment like a boss, even as it changes in real-time. In a private session with journalists, Google demoed a robot able to sort bowls of varying finishes and colors, and even discern between fake fruits like grapes and bananas, then distribute each in the appropriate bowl. Another demonstration had a robot understanding the complexity of granola in a Tupperware container that needed to be packed into a lunch bag.
At the heart of this announcement is Google praising DeepMind's work in transforming Gemini into a "brain" for the robotic sphere. It's mind-blowing to think that a brand known for powering your smartphone could one day emerges as the brains behind a humanoid robot. Carolina Parada, Senior Director and head of robots at Google's DeepMind, wrote, "We look forward to exploring our models' capabilities and continuing to develop them on the path to real-world applications."
Google is teaming up with companies like Apptronik to build the next-gen humanoid robots. The Gemini Robots-ER model will also be available for partners to test, including Agile Robots, Agility Robots, Boston Dynamics, and Enchanted Tools. The robots are on the way, but there's no timeline to worry about (at least for now).
As for those who might worry about Gemini going haywire and causing physical pain, Google assures us that their Gemini Robotics-ER models can comprehend whether or not a potential action is safe in a given context. This is achieved by basing it on the ASIMOV dataset, which helps researchers measure the safety implications of robotic actions in real-world scenarios. Google is also collaborating with experts to ensure responsible AI application development. So, buckle up and embrace the robot revolution with optimism!
The tech giant Google, with DeepMind's assistance, has announced Gemini Robotics, an advanced VLA model based on Gemini 2.0, which is designed forReal-life robots. Google's second AI model, Gemini Robots-ER, boasts enhanced spatial understanding, enabling robots to navigate their environment effectively.
Google is partnering with companies like Apptronik to create next-generation humanoid robots, and they plan to make the Gemini Robots-ER model available for partners to test. To address concerns about safety, Google's Gemini Robotics-ER models are designed to comprehend whether actions are safe in a given context, relying on the ASIMOV dataset.
Google continues to collaborate with experts to develop AI responsibly and ensure its applications adhere to high safety standards. Embrace the future as technology pushes the boundaries of artificial intelligence and drives the robot revolution.
