Google Deepmind AI knows how to take a nap

Aug. 12, 2017, 8:25 a.m. By: Pranjal Kumar

Google Deepmind

Google Deepmind is probably the most human like a machine that has been built so far by any company. It created a havoc around the world by beating top human player at the game of Go. Recently, Google also said that Deepmind can use imaginations and make a prediction just like the human does. But, the latest addition to the Deepmind's arsenal is the capability to take naps.

It might sound odd to hear first time as the main job of a computer is do take the heavy job so that humans can go to sleep. You might ask what is the benefit of the machines that needed time to sleep. But, according to the company, the capability of sleeping gives more power to the AI. This capability was the key part of the company’s deep-Q-network. Deep-Q-Network is an algorithm that learns to master much range of Atari 2600 games only with the scores and pixels as the input. AI learns from his mistake that occurred in his past in an offline mode.

The use of AI based system has increased in the recent time. From the sectors like the drone vehicles to self-driving cars, all need the heavy implementation of AI. The self-driving cars need to decide how to think like a person and how to analyze all the data. The new sleeping technique from Google means that AI can use its resources to figure out a problem and then save all the related details to dream about it later which will help it to improve its decision-making process. This will prevent Deepmind to regularly work at one problem. If it fails at something, then it can think about that problem in the offline mode and solve the problem once it is back online. This is exactly similar to the way human solve a problem.

All the resources can create a path for the creating an AI system which is exactly similar to human. We might be still very far from that but the recent progress suggests that it can be achieved.

Image Source: HitBerry