In recent years, we have witnessed notable growth both in interest and in the development of robotics. Different companies have devoted themselves to creating their own platforms, prototypes and several commercial models have already been sold, including certain humanoid design robots.
But the idea of a robot that can interact naturally with people, comply with orders and develop in unknown environments, always seems to be far away. The key to achieving it is in artificial intelligenceAnd Google hit the table announcing its own efforts on the ground. These are Gemini Robotics and Gemini Robotics-Er, two new models developed by Google Deepmind.
Google works in two new gemini versions for robotics
That a robot resembles and the humanoid conception does not mean that it is formed to imitate the behavior and the capacities of a human being. You can have hands and legs, and know how to walk, run and move objects, but today Robots work better in controlled environments. One of the biggest challenges in robotics is the real world, with its conditions that change constantly.
Something as basic as ordering the purchase of supermarkets in the kitchen is something extremely complicated for a machine, because this task is full of actions that involve products which, perhaps that, had never seen, with different shapes and characteristics, and we all know that the position and the location of objects in a kitchen of a house are very variable. This chance is something very complicated for machines.
Deepmind has developed two new models based on Gemini 2.0, which are mainly designed to improve human-robot interaction, and also for Allow robots to develop better in the real world. Gemini Robotics, a model focused on vision, language and action, allows robots to improve their response capacity in different environments and situations. Gemini Robotics-Er, on the other hand, consists of a series of tools for robotics experts.
According to Google Deepmind, in its official declaration, this new AI development for robotics has three main objectives, giving priority to the evolution of the way in which robots interact with unknown environments:
- Generality: Robots must be able to perform unexpected tasks, solve problems on the fly in new or known environments and to be able to interact with new objects, to learn at the time
- Skill: The robots must be formed with a very precise motor skills, in order to perform complex and in -depth tasks
- Interactivity: A robot must be able to contact people in natural language and also respond to their instructions, consultations and orders in real time
According to Deepmind, Gemini Robotics is designed to operate with different types of robots, not only with humanoid robots. And the company is already working with different partners to develop robots that will have these models, including Apptronik. They also work with other companies to test their new artificial intelligence, including Boston Dynamics and Agile Robotics.
Gemini Robotics is still in development, and we do not know when we can see the first robots equipped with this technology, even if they are prototypes. However, Google seems to bet on Create the android robotsAn artificial intelligence platform that can be equipped in the robots of different manufacturers, instead of developing their own equipment. At least for the moment.
Cover photo | Gemini + Arseny Togulev in Unclash (with edition)
In Xataka Android | The “godfather” of the AI predicts another technological revolution in the next five years
In Xataka Android | The Google Gemini Revolution begins with gems. What are they and how to create yours
