According to the International Energy Agency, "electricity demand from data centers worldwide is set to more than double by 2030 to around 945 terawatt-hours (TWh), slightly more than the entire electricity consumption of Japan today." However in many ways AI is also transforming the transition to greener energy practices. So what role should AI play in our energy consumption in the future?
How is AI currently helping our energy grid?
The MIT Technology Review calls the grid system as "the most complex machine ever built." This is due to the fact that our energy grid has to satisfy the ever-changing balance between supply and demand of electricity across vast geographical areas while also being managed by grid operators and organizations constantly. However this constantly flowing "machine" has become increasingly complex with the introduction of renewable energy since now the grid has to manage electricity from power plants, solar panels, and wind power sites. With so many variables to keep track of, machine learning models such as MISO (Midcontinent Independent System Operator) use mathematical equations to predict the energy demand for the coming days and also provides efficient solutions in distributing the energy to households, offices, and infrastructure.
AI's ability to breakdown a large data set of demographics and consumption trends allow it to manage our energy grid more efficiently in time and money, causing countries such as the United States to fund smart grid projects in billion dollar grants.
Do the benefits from AI implementation outweigh the environmental toll that same models cause?
Although AI is becoming a universal and accessible tool, the application of AI can:
Consume vast amount of water on a regular basis- In fact as of 2021, Google's data centers used 4.3 billion gallons of water for efficient cooling. On top of this, there are "4,165 data centers registered" and "1730 operators" across the entire United States.
Create significant carbon emissions and pollution- AI hardware requires the mining of earth minerals such as lithium and cobalt, the practice of which severly increases carbon emission and release toxic, radioactive elements to the ecosystem that surrounds mining grounds.
High electricity demand- Ironically, AI's data centers require an exceeding amount of energy and electricity in order to for its application and language learning models to be trained in analyzing and implementing mass-scale data sets. However, the presence of renewable energy in the grid can actually help offset the electricity demand that these models have since now "excess energy" is being created by solar panels and windmills, allowing for it to be properly used for AI applications.
My Take
Whether you are supportive or against the popularization of AI, we all must accept that these language learning models will be a part of our future in more ways than we might realize even now. Therefore, we need to be extremely careful with the regulation of AI applications. In terms of our energy grid, the only way that we can try to regulate the negative effects of AI operators is to fall back on the energy generated by renewable practices. However, when you take a step back, this proposal feels almost circular: to reduce the harms of AI-driven energy use, we rely more and more on renewable energy resources, which themselves are managed through artificial intelligence algorithms. As AI accelerates the transition to sustainable practices, it also becomes increasingly energy dependent, requiring ongoing innovation and rigorous oversight.
However, if we completely abolish AI from energy grid operators, how will we as humans be able to juggle the complex grid with gas and renewable energy practices being introduced at a rapid rate? Is there a capable substitute? The answer will be upsetting to some, agreeable to others: Maybe not, so we need to regulate artificial intelligence algorithms in our energy grids. Without AI, we would struggle to balance supply and demand across countries, and risk greater inefficiency or outages as electrification and renewables soar. In order to block the feedback loop, we need to implement ethical limits and sustainability goals to hold ourselves accountable. That means transparent reporting, sustainability benchmarks, and robust performance metrics for AI data centers must become industry standards—not only to minimize harms but to maximize climate benefits moving forward
Sources:
https://www.energy.gov/gdo/grid-resilience-and-innovation-partnerships-grip-program
https://insights.taylorandfrancis.com/sustainability/ai-bad-environment/#section-C8HC29mMbS
