Google has announced two artificial intelligence models to help control robots and perform specific tasks such as classification and organizing.
Gemini Robotics was described by Google as an advanced vision language action model built on Google's AI chatbot/language model Gemini 2.0. The company boasts physical behavior as a new output modality with the aim of controlling the robot.
As Google explained, gemini robotics-er means “er” meaning “er” press releasedeveloped for advanced spatial understanding, allowing roboticists to execute their own programs.
The announcement promoted the robot as being to perform “a wider range of real-world tasks” with both clamp-like robot arms and humanoid-type arms.
“To be useful and useful to people, a robotics AI model requires three main qualities. They must be general. They can adapt to a variety of situations. They must be interactive. This means that they can quickly understand and respond to environmental instructions and changes.”
The company added, “[Robots] It must be dexterous. This means that people can generally do what they can with their hands and fingers, as they manipulate objects carefully. ”
The attached video showed a robot orally praised the organisation of fruits, pens and other household items into various sections and bins. One robot was able to adapt to the environment even when the bin moved.
Other short clips in the press release include a robot playing cards or TIC-TAC-TOE and a lunch bag packing food.
The company continues, “Gemini Robotics leverages the global understanding of Gemini to generalize to new situations and solve a variety of tasks out of the box, including tasks that have never been seen before in training.”
“Gemini Robotics is also skilled in dealing with new objects, diverse instructions, and new environments,” Google added.
What they haven't said
Telsa Robots displayed similar features near the start of 2024.
Google did not explain to readers that this is not a new technology. Also, the innovation is particularly impressive given what is already known about Advanced Robotics.
In fact, it was in mid-2023 that a group of Princeton University scientists and robotics engineers exhibited robots that allow them to learn personal cleaning habits and techniques to properly organize their homes.
If necessary, the bot can also throw away the trash.
For “Tidybot”, I entered a user who explained the sample settings to tell the robot where to place the items. Examples such as “A yellow shirt goes to the drawer and a dark purple shirt goes to the closet” were used. The robot summarises these language models and supplements the database with images found online, allowing it to compare objects and images in the room to identify exactly what it was looking for.
The bot was able to fold laundry, put trash in boxes and organize his clothes in various drawers.
About six to seven months later, Tesla reveals a similar technique when he shows the robot “Tesla Optimus”, removing the T-shirt from the laundry basket and then gently folds it onto the table.
Essentially, Google seems to simply allow robot voice-to-text commands, rather than connecting its language model to existing technology and entering commands via text alone.
Like Blaze News? Bypass censorship, sign up for our newsletter and get stories like these directly into your inbox. Sign up here!
