Google’s powerful DeepMind AI is saving the company a lot of dough. The technology has resulted in a 15 percent improvement in power usage efficiency (PUE). Get your calculators out. Google said it used 4,402,836 MWh of electricity in 2014 which is the equivalent to 366,903 family homes in the United States. The average American family paid between $25 to $40 USD per MWh in 2014.
If we split this difference, we can theorize that Google spent at least $140,890,752 just on its 2014 electric bill. A fifteen percent improvement means a little over $21 million in savings. Most of Google’s electric bill comes from powering its large servers.
In 2014 the company said it used neural networks to predict how its power usage would change over time. The DeepMind AI system, which is being used as a replacement, controls parts of its data centers to reduce power consumption by manipulating computer servers and related equipment like cooling systems. The software changes how a system runs by trying to to get “the highest score” as if the system was a computer game. In this situation, a higher score means more efficient consumption of electricity.
DeepMind Co-Founder Demis Hassabis remarked, “ [DeepMind] controls about 120 variables in the data centers. The fans and the cooling systems and so on, and windows and other things. [We] were astounded.”
DeepMind has recently been implemented in other projects. Google recently commenced a new collaboration with the United Kingdom’s National Health Service (NHS) in order to create a machine learning system that will be able use digital eye scans of the eye to recognize sight-threatening conditions. At Google headquarters, DeepMind will now try to put in more sensors so that they can make the building even more energy efficient.