Using DeepMind artificial intelligence technology, Google Province has a large bill for electricity bills

When Google bought DeepMind at a high price of 400 million pounds (about 600 million U.S. dollars), many people couldn't understand it, and even wondered why it would acquire the unnamed start-up company at that time. Now it's the turn of those who didn't look good on DeepMind in the past. Using DeepMind artificial intelligence technology, Google Data Center has saved a lot of electricity expenses.

In recent months, Google has begun deploying its DeepMind artificial intelligence system in its data center to control and reduce the power consumption of computer servers and related equipment (such as air-conditioning systems) in real time. It is reported that DeepMind software uses the same AI technology to "self-training" video games.

Although the DeepMind artificial intelligence system only helps the data center to reduce the power consumption by several percentage points, it also saves a lot of costs due to the huge amount. DeepMind, on the other hand, has helped Google’s data center usage increase by 15 percentage points. According to 2014 figures, Google’s total power consumption reached 4,402,836 MWh, which is equivalent to the average annual power consumption of 366,903 US households. The largest power consumption among them is the data center – a webpage that supports global use. And mobile application services.

Saving money is equivalent to making money

Technology giants like Google, which use a few percentage points of electricity, mean that they save a lot of money. According to the data from the US Energy Information Administration, the electricity price of industrial electricity in the United States is about 25-40 US dollars per megawatt-hour. For example, if Google’s data center power consumption is reduced by 10%, it will help them save for the next few years. Billions of dollars spent.

Google had previously stated that they will apply DeepMind's machine learning technology (an artificial intelligence type) to its data centers. In fact, as early as 2014, DeepMind developed a neural network (a model identification system) that began to predict the power usage of Google data centers in the coming years.

Of course, Google's acquisition of DeepMind is not intended to train them to play video games. Using artificial intelligence technology, DeepMind software can change the operating parameters of data center internal equipment according to the environment, and then get the best performance and the lowest power consumption.

Demis Hassabis, co-founder of DeepMind, said:

DeepMind artificial intelligence software controls about 120 device parameter variables inside the data center, including fans, air conditioning systems, even windows, and so on. From the use of the situation, the effect is very good. Now this energy-saving project is just beginning, but DeepMind already knows that this artificial intelligence solution can play a role. In the next step, Google will deploy more sensors inside the data center to collect data and help this artificial intelligence software to function more efficiently.

VIA datacenterknowledge

This entry was posted in on