AI technology has just recently been on the pages of newspapers just in connection with the victory of the Deep Blue computer over the world’s chess champion Garri Kasparov. Now AI penetrates all sectors including energy. However, their operation is energy-intensive and the question arises whether the very essence of AI does not create a dangerous conflict of interest in its use in energetics.
Greentech Media reported, in the US energy storage market they expect to double twice this year over the 2013-2017 period. It will also be through AI technologies.
For example, Californian company Stem has developed the Athena system. AI is used for power consumption mapping. It allows customers to track energy-price fluctuations and make more efficient use of electricity storage. It has earned $ 374.5 million from a diverse group of investors, including the US Department of Energy.
Government support has also been provided by the Stanford University team. This supports the development of a system that, through artificial intelligence, can independently assess the risks of accidental events in a power grid and perform interventions to stabilize it.
Google has also created Sunroof, which can evaluate and manage the use of solar energy. It works on the basis of weather, electricity prices, and other parameters. And German Siemens has released a software package that uses, for example, electromobility charging for active network management. In coal power plants, they test the SparkCognition tool for failures. Chevron or BP Ventures are starting to use AI technology to find new oil or gas fields.
Artificial intelligence saves and consumes
As the use of AI technologies grows, their consumption is also increasing. In 2014, data centers around the world consumed 352 TWh of electricity. US data centers have contributed 70 TWh. That is the same as the annual consumption of more than six million average US households, reported Menerga company.
The entire Czech Republic consumed 68.5 TWh in 2014 and less than 74 TWh in 2017.
Last year, the total consumption of world data centers was already 428 TWh. This is 21 percent more than in 2014. This year it could be an increase of 6 percent. Data centers around the world are already consuming significantly more power than the UK. Their share of global energy consumption exceeded 3 percent. They expect it to treble over the next decade.
It is from the nature of the functioning of AI that there is a danger that is being raised by more and more experts. AI has the ability to learn and improve on the basis of both observed. There is no limit for internal data by any predefined set of possible situations. Therefore, its decision-making must not, in principle, be determined solely by the interests of man. One can therefore simply imagine a situation where AI chooses to prioritize its own energy needs in order to maintain its functioning.
Microsoft’s Jeff Bullwinkel at a recent conference called Mankind at Risk: AI and changing societies have highlighted the fact that the consumption of technology giants such as Google or Microsoft may correspond to the consumption of a medium-sized European country over the next decade, Deutsche Welle said.
According to Professor Ian Bitterlin of Leeds University, Japanese data centers would have to consume all the energy produced in Japan by 2030, if their consumption growth continues at the same rate as before, according to The Independent. A distinguished American AI experts believe that the only way is to change the computational architectures and artificial intelligence practices, which inevitably appear to be due to the enormous demand for data center services.
It is almost certain that AI will take over the control of energy systems in the near future. It is not possible to predict precisely the direction in which the AI will evolve. The extent to which it will be autonomous and what decision-making space it will take.
However, it is obvious that one of AI’s significant motivations to minimize energy expenditure may be its own consumption. Although there is no imminent risk now, in the longer term, human staff should be mindful of the conflict of interest that AI naturally gets or gets. The vision of a more or less distant future, when AI begins to optimize the “human processes” itself because of its own energy demands and needs, is certainly not a concept that would please anyone.