After the acquisition of DeepMind by Google, they have also started using their technology to better their business. DeepMind Co-Founder Demis Hassabis revealed that Google is using technology from the DeepMind artificial intelligence for cutting down on their power consumption by its data centers.
The Alphabet Inc. unit has put a DeepMind AI system recently which controls the parts of its data centers and reduces the power consumption by manipulating computer servers and related equipment such as the cooling systems etc. The self learning program uses a similar technique to the previous DeepMind software which taught itself to play Atari video games.
The result is an amazing savings of up to 15 percent when it comes to power usage efficiency, or PUE. PUE is a unit which measures how much electricity Google uses for its computers, versus the supporting infrastructure like cooling systems.
As per an official statement Google revealed that is takes up as much electricity as an average of about 366,903 house hold homes and uses precisely 4,402,836 MWh per year. And since the data centers consume the highest percentage of electricity, installing the DeepMind AI system was only logical.
A cut down on electricity of this magnitude translates into major financial profits. Electricity prices in the U.S. range from about $25 to $40 per MWh, according to data from the U.S. Energy Information Administration. And this means saving 15 percent on data center power consumption saves hundreds of millions of dollars over multiple years.
DeepMind’s technology is built on the previous such efforts by Google in order to apply machine learning to its data centers. In 2014, the company used neural networks which are a type of pattern recognition system, in order to predict how its power consumption changes over time in a bid to increase the efficiency of the equipment.
The DeepMind work is going even further by applying similar logic to what it did with the Atari games. Instead of playing the game, the software changes how equipment runs inside the data centers in order to get the highest score which translates into more energy efficiency.
The algorithm controls and monitors about 120 variables in the data centers, like the windows, the number of people, the cooling systems etc. And now knowing that the system works, the company aims at providing more data to its system by installing more sensors in the Google data centers. This will translate into more electricity efficiency for Google, and thus more profits!
Do you have anything more to add to this article? Comment below!