Another day, another awesome implementation of AI from Google. To combat the notorious energy consumption of data centers worldwide, Google has been using an AI system built by DeepMind to control certain parts of its data centers for increased environmental friendliness. For those who haven’t seen our article about AlphaGo, another DeepMind project by Google, here is a quick link to check it out. Although DeepMind is famous for developing AI technology and was able to beat the world’s best Go player, it hasn’t turned a profit for Google. Yet.
The DeepMind AI system has been used to reduce the energy consumption of cooling units in Google data centers. These coolers are vital as they prevent overheating and so far DeepMind has reduced their energy costs by up to 40 percent. How? By training DeepMind’s self-learning algorithms to predict hourly heat patterns. The coolers were then able to run at the most efficient intensities to keep the servers sufficiently cool. It’s as simple of a concept as a weatherman telling you whether you’ll need a jacket or a T-shirt to prevent temperature discomfort but for a machine to regulate this in perfect sync with thousands of other machines is an absolute marvel. Using a system of neural networks that analyze different operating scenarios, DeepMind makes a more efficient framework to understand data center dynamics.
As stated by DeepMind reasearcher Jim Evans, “It’s all about optimization. It can be used in supply logistics, shipping logistics, and dynamic pricing so we’ll be seeing AI applied in a lot more areas.” Google is continuously on the forefront of AI development and exploration so as with the AlphaGo project, this opens up a variety of doors and directions. DeepMind and their machine learning techniques will no doubt make an appearance with their projects later this year and as always, we’re excited to see what application comes next.