Google’s DeepMind AI division has unveiled a self-improving robot that can teach itself new tasks without human supervision. This product is called RoboCat and DeepMind claims that this robot is the first model in the world that can learn and solve various problems by modeling differe robots in the real world.
nnnn
According to Hoshio, by designing this robot, DeepMind has registered a great progress towards creating general robots, capable of performing millions of daily tasks.
nnnn
When the robot performs an action, it generates data that the AI can use to improve its technique. This advance in technique is not limited to robots that generate data, but can be transferred to other robotic systems as well.
nnnn
DeepMind Technologies, the London-based company that Google bought in 2014, said: “This technology represes a significa advance in the field of building general-purpose robots that can perform everyday tasks.”
nnnn
In a blog post detailing their latest AI model, DeepMind researchers wrote: “RoboCat learns much faster than other advanced models. This bot can perform a new task in less than 100 visits because it uses a large and diverse data set. “This importa capability will help accelerate robotics research by reducing the need for human-supervised training and is an importa step toward creating a general-purpose robot.”
nnn
nnn
RoboCat is inspired by DeepMind’s artificial ielligence model Gato, which learns by analyzing text, images, and eves.
nnnn
The researchers trained RoboCat by showing it tasks such as fitting shapes io holes and picking up pieces of fruit that a human-corolled robot arm would perform.
nnnn
RoboCat then trained on its own and was able to consistely improve, as it had done this task an average of 10,000 times without human supervision.
nnnn
During the DeepMind experimes, this artificial ielligence robot trained itself to perform 253 tasks in four differe types, and during the training it was also able to adapt and improve its abilities.
nnnn
According to researchers, AI has the poteial to learn things and learn how to do them that it has never even been exposed to before.
nnnn
“RoboCat has a learning cycle such that when it learns new tasks, it performs better at learning other new tasks,” a blog post said.
nnnn
These improvemes show that RoboCat has the poteial to quickly and efficiely adapt to a wide range of tasks and environmes, making it a powerful tool in the field of robotics. Something similar to how people develop a more diverse range of skills as they coinue to learn and deepen their knowledge in a particular area.
nnnn
The research follows a growing trend of self-learning robotic systems, such systems have the poteial to create robots that are able to learn and adapt to new situations, something akin to the concept of domesticated robots in science fiction.
nnn
nntt
