Artificial intelligence helps Google to save electricity Bashny.Net

One of the main and the main items of expenditure Google Inc. - electricity. Data centers consume a lot of the search giant's energy, while constantly heating up and releasing heat, thus requiring them still and cool, but it also leads to energy costs. Otherwise, the server may be damaged, and it is also costing a pretty penny. Therefore, in order not to consume electricity in vain, as well as improve the efficiency of its facilities, the corporation a few years ago decided to start something to do, here and useful new startup DeepMind, working on the creation of artificial intelligence, which Google on the occasion and bought.




Previously, experts DeepMind taught artificial intelligence quite quickly chopped video games on the Atari console, later developed a program to beat the champion of the world board game Go, and now guys are not without their developments to help Google to save money. Who is under the wing of the Alphabet, Google parent company, the project brings certain benefits.

In DeepMind we created a neural network system and connect it to the data center. Additionally, included in the circuit program, able to predict the temperature based on the data obtained previously. The system became rather complicated, it works with a variety of parameters, at the same time taking into account the speed of cooler, the efficiency of cooling systems in general, windows and more.



Artificial intelligence is used by Google for a long time, with the servers it too "friendly" for a couple of years, but developed DeepMind program helped make the work more efficient data centers. As stated by representatives of DeepMind, electricity costs required to cool their own servers, declined by almost 40 percent. Indicators measured by Google, more modest, but still encouraging.

The graph above shows the power consumption before the system was launched DeepMind and after its launch. The right side of the chart shows how the performance fell off after repeated it.