Google is trusting an artificial intelligence (AI) system developed by DeepMind to stop its data centres around the world from overheating.
The AI system — able to reduce the amount of energy Google used to cool its data centres by 40% — has been giving cooling recommendations to Google’s data centre operators since 2016.
But now Google is allowing the data centre operators to take a back seat, giving the AI an unprecedented level of autonomy in the process.
“In 2016, we jointly developed an AI-powered recommendation system to improve the energy efficiency of Google’s already highly-optimised data centres. Our thinking was simple: even minor improvements would provide significant energy savings and reduce CO2 emissions to help combat climate change,” wrote Google data centre engineer Amanda Gasparik, DeepMind research engineer Chris Gamble, and DeepMind team lead Jim Gao in a joint blog post published on DeepMind’s website.
“Now we’re taking this system to the next level: instead of human-implemented recommendations, our AI system is directly controlling data centre cooling, while remaining under the expert supervision of our data centre operators. This first-of-its-kind cloud-based control system is now safely delivering energy savings in multiple Google data centres.”
DeepMind’s deep neural networks look at data from the cooling systems in Google’s data centres every five minutes to predict how different combinations of potential actions will impact future energy consumption.
“The AI system then identifies which actions will minimise the energy consumption while satisfying a robust set of safety constraints,” the blog post reads. “Those actions are sent back to the data centre, where the actions are verified by the local control system and then implemented.”
Around 2% of global greenhouse emissions are produced by data centres, which is about the same amount as air travel. Google has over 15 of the most energy efficient data centres in the world but it wants to further reduce their carbon footprint.
DeepMind believes that its technology could potentially be used to reduce the energy consumption for other large pieces of infrastructure, including the National Grid in the UK.
“What we’re now thinking about is how we can extend this more broadly to the national grid and to other large-scale pieces of infrastructure which essentially have the same characteristic,” said DeepMind cofounder Mustafa Suleyman at the Wired 2016 conference.
“All of our algorithms that we develop are inherently general so given some data set, we should be able to train an algorithm based on some inputs, develop a model, predict some outputs, and then, providing we have access to some controls, we should be able to deliver similar sorts of performance.”
Several months later, Demis Hassabis, CEO and cofounder of DeepMind, confirmed that DeepMind had been in talks with the National Grid. He told The Financial Times that it would be great if 10% of the UK’s energy usage could be cut without any new infrastructure.
“We’re early stages talking to National Grid and other big providers about how we could look at the sorts of problems they have,” Hassabis said. It would be amazing if you could save 10% of the country’s energy usage without any new infrastructure, just from optimisation. That’s pretty exciting.”