In 2014, Google acquired DeepMind, an artificial intelligence company, for US$600 million. If Alphago earned the deal a good reputation, then Google recently started earning back the principal. According to Bloomberg News, DeepMind co-founder Demis Hassabis recently stated that the company's artificial intelligence technology has been used to save data centers energy consumption.
It is understood that Google has used DeepMind's artificial intelligence system in control data centers in recent months to save energy by controlling computer servers and cooling systems. The technology used was previously used to teach computers to play Atari games.
Hassabis said that the entire system has reduced power usage by several percentage points, which is a good thing for both cost and environment. Similarly, energy efficiency has increased by 15%.
In 2014, Google’s electricity consumption was 4,402,836 MWh, and a large part of it was consumed by data centers. These centers are also the foundation for company services. Saving a few percentage points of electricity is a big deal for Google’s finances. When U.S. companies use electricity at a price of $25-40 per megawatt hour, a 10% reduction in power consumption means savings of hundreds of millions of dollars in expenditure within a few years. Google’s acquisition of Deepmind used 600 million U.S. dollars.
However, Google did not think about artificial intelligence when it had DeepMind. It used machine learning technology in a data center in 2014. At that time it used neural networks to predict changes in energy consumption over time, thereby arranging equipment usage more efficiently.
DeepMind's role goes even further. Unlike the number of steps in the game, the software will change the way the data center equipment operates to get the highest score, which means that it is more energy efficient. Hassabis said that the system controls 120 variables in the data center, including fans, cooling systems, and windows.
This is just the beginning. Knowing that this approach works, DeepMind also learned what information the artificial intelligence system still lacks, and will allow Google to place more sensors in the data center to make the software more efficient.