Google says DeepMind AI reduces data centre energy consumption by 30 percent


Google has 15 data centres around the world to help manage services like Google Search, YouTube, Gmail and more.

However, with so many tasks to be carried out on a daily basis, there is also a large amount of energy consumption taking place in each of these centres.

With that in mind, Google revealed two years ago that it had started using its DeepMind AI to bring about more efficient operation of its data centres. Specifically, Google was able to teach DeepMind AI to read thousands of sensors in a data centre and predict which cooling system actions are needed at any given time, all through machine learning. This led to energy consumption reduction of 30 percent on average.

Now, Google says it’s taking this process one step further. While DeepMind AI would previously use human-implemented recommendations, the AI system is now responsible for directly controlling data centre cooling.

Google noted that centre operators will still supervise the procedure in case any issues arise. However, automating the system will help operators “implement more granular actions at greater frequency, while making fewer mistakes,” according to data centre operator Dan Fuenffinger.

Going forward, Google says it sees potential to apply this technology in other industrial settings beyond data centres.

Source: DeepMind