Google’s DeepMind AI Helps Conserve Energy

Google uses DeepMind AI

Tech companies have always felt that energy consumption by big data centers is a problem that must be addressed. Server cooling has been a big problem for many of the big data companies, including Facebook. To resolve the issue, Facebook made one of its facilities on the edge of the Arctic Circle. Google, however, decided to employ a bigger and much better solution. The tech giant employed DeepMind Artificial Intelligence Unit as the in charge of the facility. The AI was responsible for managing the power usage across different parts of the data centers. As a result, the company managed to cut down the electricity consumption of server cooling by an astonishing 40 percent. Google has described this move by stating that it is a ‘phenomenal step forward’.

Google stated that after putting in account the electrical losses and cooling inefficiencies, this 40 percent power saving translated into roughly 15 percent reduction in overall consumption. Considering the astonishing amount of electrical energy consumed by the company’s data centers (4,402,836 MWh, back in 2014, which is roughly equivalent to the energy consumption of at least 366,903 US households), this reduction of 15 percent will equate to savings of millions of dollars over the years. Plus, when you consider the company’s acquisition of UK-based DeepMind for $600 million, back in 2014, then it seems that the company’s acquisition will pay off in a few years down the road.

DeepMind Co-Founder, Demis Hassabis stated in an interview with Bloomberg that the neural networks designed by the company control over 120 variables of the data centers which include components like fans and cooling systems. When put in charge, the system analyzed the data given out by sensors placed on server racks (which record basic information like pump speeds and temperature) and after a complete analysis worked out the most effective way to cool the servers. The scientists behind the project stated that the next step is to work out the components that can provide more data for calculating efficiencies and place sensors around them.

The best part is that the company has no intention of stopping at Google’s data centers. Instead, the company stated that since it is a general-purpose algorithm that can crunch complex dynamics, Google intends to employ it across other environments.

No votes yet.
Please wait...
Voting is currently disabled, data maintenance in progress.