DeepMind’s new Gato AI makes me fear humans will never achieve General AI

DeepMind today unveiled a multimedia artificial intelligence system capable of performing more than 600 different tasks.

It’s called Gato, and it’s arguably the world’s most impressive all-in-one machine learning toolkit to date.

According to DeepMind Blog post:

The agent, which we refer to as Gato, operates as a multi-modal, multi-tasking, multi-incarnation public policy. The same network with the same weights can run Atari, pictograms, chat, stack blocks with a real bot arm and much more, and decide based on their context whether to output text, shared torque, button presses or other tokens.

A diagram showing some of the many functions that Gato can perform.