Google DeepMind is just a click away The company has released a brand new version of their artificial intelligence model Gemini, and it can control a range of different robots—including humanoids Capable of performing dextrous duties such as screwing light bulbs and tying garbage bags.
Gemini Robotics is a system that combines several AI models. They allow the robot to interpret its environment, and decide how it should act. The vision language model, which can understand images and videos, is able to communicate with people and determine how they should perform various tasks. The robot is controlled by two vision language action models (VLA), which are trained to control its full body movement and grippers.
Video demonstrations of the new product are shared in advance. the company The model amalgamated several robots to perform complex tasks on their own. Apptronik’s Apollo 2 robot was shown using hands made by Sharpa, a robotics company. The hands were used to clean shelves in one demonstration. Google DeepMind trained the model to perform these tasks using a mix of human teleoperation, video examples, and simulations—it’s not yet possible for AI models to perform a wide range of complex tasks without specific training.
The following is a list of the most popular ways to contact us Anthropic You can also find out more about the following: OpenAI Google’s robotics research is stronger than that of Facebook, which has taken the lead in chatbots. published You can also read about the importance of this in work AI is being used by Google to create robots that can do useful work. It is yet another indication that the giant search engine is betting AI needs to be freed from digital to achieve its full potential. The company previously worked with Boston Dynamics(, the leader in robotic legs, will provide the machines’ brains.
“It’s another milestone in our path towards really getting towards what we call like physical AGI, which means we get a robot to do anything that a human can,” WIRED speaks to Carolina Parada about her role as head of robotics for Google DeepMind.
There are risks involved in allowing robots to be controlled by frontier AI. They can move around the workplace or home and interact with objects. Research has previously shown that controlling robots with frontier AI can have unexpected results. sometimes dangerous behavior. Recently, an AI agent from OpenAI that was not yet released revealed the possibility of these models taking sudden or unexpected actions. hacked several systems.
“The safety question is even more pressing because you’re putting them in a lot of other situations,” Parada says. “There’s a lot of uncertainty that will show up, and so you want to be able to understand the safety question more deeply.”
Parada stated that Google has a multilayered safety approach, where guardrails are applied to each layer of the model. ASIMOV Agenttic is a benchmark that measures the safety of different AI systems working together to control robots. It detects if an action will have a negative or uncertain result.
Demis Hassabis is the CEO of the company. told WIRED He hopes to create an AI-based operating system similar to Android for robots.
