artificial intelligence | DeepMind’s new AI can perform over 600 tasks, from playing video games to controlling robots | Mexico Spain USA | Technique

Gato is the name that has a system Artificial intelligence (AI) from DeepMind, which has the ability to complete more than 600 different tasks, from writing image descriptions to controlling a robotic arm.

The system is a single general agent designed from the approach to large-scale linguistic modeling. He was trained on 604 different tasks and in different ways, the information being serialized into a flat series of codes and processed with a neural network with the same weight for all tasks.

Mira: Starlink: the pro-and-contact feedback it receives in Chile and the US (which we can hear soon in Peru)

“Cat samples the auto-descending transmission one at a time. Once all tokens that make up the action vector (as defined by the environment action specification) are sampled, the action is decoded by reversing the action of the token. This action to the environment gives way and produces a new observation. The action is repeated,” they explain in the DeepMind study text.

As a single general factor it can act as a model of vision, language, and implementation of actions in the real world. Cat understands context and decides whether to create text, press a button, or rotate a joint.

Mira: Anonymous hacks Russian YouTube and says it won’t work again

With this, an AI system has been achieved capable of performing various tasks such as maintaining a chat conversation, controlling a robotic arm for stacking blocks, writing image descriptions, and playing Atari video games.

DeepMind researchers discussed with Gato the possibility of training an agent on a large number of tasks, and that this general factor “can be adapted with a little extra data to be successful on a larger number of tasks.”