Google Gemini Robotics To Bring GenAI To Real World Robots ...
The trend of artificial intelligence previously moved away from robotics, and it looks like it’s heading back there again. OpenAI has at least considered it, and prior reports indicate that Apple is actively looking into it. More recently, Google has announced its own initiative in the same direction, with the announcement of Gemini Robotics.
Google's Approach to AI in Robotics
Via its AI division DeepMind, the internet search giant says that “in order for AI to be useful and helpful to people in the physical realm, they have to demonstrate ’embodied’ reasoning”. This is described as “the humanlike ability to comprehend and react to the world around us”. Of course, there’s the need for the robot in question to “safely take action to get things done”.
Gemini Robotics and Robotics-ER Models
Google says that Gemini Robotics is “advanced vision-language-action (VLA) model that was built on Gemini 2.0 with the addition of physical actions as a new output modality for the purpose of directly controlling robots”. There’s a second model as well, dubbed simply Robotics-ER, which gets the additional “advanced spatial understanding” as well as embodied reasoning (ER) that gives it its name.
Principal Qualities of Google's Robots
Overall though, Google has three principal qualities that it wants its robots to have. The first being Generality, or the ability for it to be “adept at dealing with new objects, diverse instructions, and new environments”, including “tasks it has never seen before in training”. Then there’s Interactivity, which lets it “understand and respond quickly to instructions or changes in their environment”. On that first point, the robot must also be able to “respond to commands phrased in everyday, conversational language and in different languages.
Finally, Google wants its Gemini Robotics robots to have Dexterity, or the ability to “do the kinds of things people generally can do with their hands and fingers, like carefully manipulate objects”. Examples include “origami folding or packing a snack into a Ziploc bag”.
Future of Gemini Robotics
For now, it’s probably too early to think about what can come out of this announcement. But for what it’s worth, the company says that the “Gemini Robotics-ER model is also available to trusted testers including Agile Robots, Agility Robots, Boston Dynamics, and Enchanted Tools”. So there are plenty of testers to help iron out wrinkles, and maybe accelerate the making of a consumer product, as unlikely as it is.
(Source: Google)
