The latest trick Google has meticulously researches on engine spin is to make AI sleep. Google wants to make AI more humane. Therefore, they now teach DeepMind to sleep.
"At first glance, building artificial intelligence that sleeps may sound unimportant AI should be able to continue to solve computing problems even after its programmers go to bed," Google wrote in a blog post.
"However, this principle is an important part of the deep-Q (DQN) network, an algorithm that learns how to master various Atari 2600 games with pixels and values as inputs."
DeepMind researchers are now teaching computer a way to learn. Words such as neural networks, machine learning algorithms, and artificial intelligence basically mean the process of teaching machines so that he can solve problems independently.
For example, to make an autonomous car work, it must have the ability to make traffic-related decisions, the data analysis algorithm should be able to group the various information acquired into the segments and the AI should be able to think like humans.
The method Google attempts now means, when a computer uses all its resources to solve a problem, it will keep the information to be solved when it is offline.
A computer does not have to be in an active state to solve the problem. If the computer fails to solve a problem, it will then disable itself, and it will be able to resolve the problem once it wakes up.