homehome Home chatchat Notifications


Google used DeepMind to cut their electricity bill by a whopping 15%

That adds up to a lot of money.

Alexandru Micu
July 20, 2016 @ 10:07 pm

share Share

Google is putting DeepMind’s machine learning to work on managing their sprawling data centers’ energy usage, and it’s is performing like a boss — the company reports a 15% drop in consumption since the AI took over.

Image via brionv/flickr

Google is undeniably a huge part of western civilization. We don’t search for something on the Internet anymore, we google it. The company’s data servers pretty much handle all of my mail at this point, along with YouTube, social media platforms and much more. But even so, it’s easy to forget that the Google we know and interact with every day is just the tip of the iceberg; it relies on huge data servers to process, transfer and store information — and all this hardware needs a lot of power.

So much power, in fact, that the company decided to do something about it. On Wednesday, Google said it had proved it could cut the energy use of its data centers by 15% using machine learning from DeepMind, the AI company it bought in 2014. These centers use up significant power to cool and maintain an ideal working environment for the servers — requiring constant adjustments of air temperature, pressure, and humidity.

“It’s one of those perfect examples of a setting where humans have a really good intuition they’ve developed over time but the machine learning algorithm has so much more data that describes real-world conditions [five years in this case]” said Mustafa Suleyman, DeepMind’s co-founder.

“It’s much more than any human has ever been able to experience, and it’s able to learn from all sorts of niche little edge cases seen in the data that a human wouldn’t be able to identify. So it’s able to tune the settings much more subtly and much more accurately.”

Suleyman said that the reduction in power use was achieved through a combination of factors. On one hand, DeepMind is able to more accurately predict incoming computational load — in other words, it could estimate when people accessed more data-heavy content such as YouTube videos. The system also matched that prediction more quickly to the required cooling load than human operators.

“It’s about tweaking all of the knobs simultaneously,” he said.

Ok, so Google’s electricity bill just went down; good for them, but what does this have to do with us? Well, a lot, actually. Data centers gobble up a lot of energy, and that means a lot of greenhouse gas emissions — combined, data centers have emission levels similar to those seen in aviation. When Google first disclosed its carbon footprint in 2011 it was roughly equivalent to Laos’s annual emissions but since then they claim they upped their game, getting 3.5 times as much computational power for the same amount of energy. Using machine learning is only the latest step in optimizing their system. The company began toying with this idea two years ago, and since then they’ve tested it on “more than 1%” of its servers, Suleyman said. It is now being used across a “double-digit percentage” of all Google’s data centers globally and will be applied across all of them by the end of the year.

Using machine learning is only the latest step in optimizing their system. The company began toying with this idea two years ago, and since then they’ve tested it on “more than 1%” of its servers, Suleyman said. It is now being used across a “double-digit percentage” of all Google’s data centers globally and will be applied across all of them by the end of the year. They haven’t released the exact amount of power their data centers use, but claims that in total its activity makes up 0.01% of global electricity use (and most of that probably goes towards the data centers.)

But DeepMind is leaving a considerable mark on their energy efficiency. It cut energy expenditure for cooling by 40%, which reduced the company’s overall power consumption by 15%.

“I really think this is just the beginning. There are lots more opportunities to find efficiencies in data centre infrastructure,” Suleyman added.

“One of the most exciting things is the kind of algorithms we develop are inherently general … that means the same machine learning system should be able to perform well in a wide variety of environments [think power generation facilities or energy networks].”

Sophia Flucker, the director of Operational Intelligence, a UK-based consultancy that advises data centers on their energy use, said it was feasible that Google had achieved such a big reduction.

“I’ve worked with some award-winning data centres, which still had plenty of room for improvement,” she said.

 

share Share

This Film Shaped Like Shark Skin Makes Planes More Aerodynamic and Saves Billions in Fuel

Mimicking shark skin may help aviation shed fuel—and carbon

China Just Made the World's Fastest Transistor and It Is Not Made of Silicon

The new transistor runs 40% faster and uses less power.

Ice Age Humans in Ukraine Were Masterful Fire Benders, New Study Shows

Ice Age humans mastered fire with astonishing precision.

The "Bone Collector" Caterpillar Disguises Itself With the Bodies of Its Victims and Lives in Spider Webs

This insect doesn't play with its food. It just wears it.

University of Zurich Researchers Secretly Deployed AI Bots on Reddit in Unauthorized Study

The revelation has sparked outrage across the internet.

Giant Brain Study Took Seven Years to Test the Two Biggest Theories of Consciousness. Here's What Scientists Found

Both came up short but the search for human consciousness continues.

The Cybertruck is all tricks and no truck, a musky Tesla fail

Tesla’s baking sheet on wheels rides fast in the recall lane toward a dead end where dysfunctional men gather.

British archaeologists find ancient coin horde "wrapped like a pasty"

Archaeologists discover 11th-century coin hoard, shedding light on a turbulent era.

The Fat Around Your Thighs Might Be Affecting Your Mental Health

New research finds that where fat is stored—not just how much you have—might shape your mood.

Astronauts May Soon Eat Fresh Fish Farmed on the Moon

Scientists hope Lunar Hatch will make fresh fish part of space missions' menus.