Google Data Centers Use AI to Cut Energy Usage
Who remembers the movie, I Robot? More than a few years old, this movie took us down the path of potential danger when artificial intelligence (AI) is given too much power in a human-dominated world. Not the first sci-fi approach to the concept of AI and what it can mean for our future, but you can’t help but like something with Will Smith at the center.
Now, Google is using AI as a way to help cut the amount of energy being used in its data centers. Unfortunately, humans aren’t getting the job done. According to this piece from The Guardian, the Internet giant opted for AI as it’s more efficient than human control. This is one example of transforming network infrastructure that is truly affecting the outcome.
The reduction in energy usage is significant if you consider that the servers that currently power streamed films, web searches and social media accounts are responsible for roughly 2 percent of global greenhouse gas emissions. Not surprisingly, Google is believed to have one of the biggest server fleets in the world. Now, those servers will be managed by DeepMind, a British AI company Google purchased in 2014.
According to Mustafa Suleyman, co-founder of DeepMind, the level of complexity and the number of variables involved in the management of a data center made it optimal for the use of AI. An algorithm, according to Suleyman, could outperform a human. While humans have developed considerable knowledge and intuition over time, the machine-learning algorithm has five years of real-world conditions data to help drive decisions.
The focus on transforming network infrastructure to produce better outcomes in energy reduction was accomplished through a combination of DeepMind more accurately predicting the incoming load and then very quickly matching that information to the cooling load required. As the environmental impact of the online world has come under intense scrutiny in recent years and Google garners so much attention in terms of data management, it makes sense that the giant would want to test different methods to reduce their environmental impact.
“I really think this is just the beginning. There are lots more opportunities to find efficiencies in data centre infrastructure,” said Suleyman in The Guardian. “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 [such as power generation facilities and energy networks].”
Given the success of this test, we’ll likely continue to see AI used to make other environments more efficient. Should Will Smith be worried that his movie plot is coming to life? Let’s hope that fantasy remains a fantasy.
