Is Artificial Intelligence A Net-Positive For Carbon Emissions?

By Haley Zaremba

-- Artificial Intelligence is already playing a major role in the energy industry with smart grids, efficiency improvements, and helping to locate new energy resources and sites. -- Artificial Intelligence requires vast amounts of energy to fuel the training and machine learning processes that make the model useful, meaning it could add to the carbon problem. -- Ultimately, AI can be a net positive if used responsibly and efficiently, but it needs to be implemented in a manner that is not wasteful and that considers its energy sources. Artificial Intelligence is increasingly becoming an essential component of the energy industry. As world leaders get more serious about meeting climate goals, the energy industry is facing the mandate to completely transform the way it operates at an unprecedented scale which will require massive, complex and nuanced computing power. AI is already playing a major role in renewable energy forecasting, smart grids, coordination of energy demand and distribution, maximizing efficiency of power production, and research and development of new materials.

A 2021 explainer from the World Economic Forum laid out three key driving factors which are “huge strategic and operational challenges to the energy system and to energy-intensive industries,” thereby making AI an essential component of the energy transition:

1. The almost unfathomable scale of the energy transition required for rapid decarbonization: in the energy sector alone, reaching net-zero greenhouse gas emissions will require infrastructure investments costing between $92 trillion and $173 trillion of by 2050, according to estimatesby BloombergNEF. AI has a massive role to play here, as “even small gains in flexibility, efficiency or capacity in clean energy and low-carbon industry can therefore lead to trillions in value and savings.” 2. A changing power sector. Electricity is overtaking fossil fuel-powered energy, creating new demand for highly complex computations for “forecasting, coordination, and flexible consumption” which are far beyond the capabilities of traditional grids. This is made all the more complex by the variability of renewable energies like wind and solar, as well as the changing producer-consumer relationship created by decentralized power production through solar panels. 3. Distribution and decentralization. New demands on the grid are made all the more complex by the variability of renewable energies like wind and solar, as well as the changing producer-consumer relationship created by decentralized power production through solar panels. Decarbonization is increasingly driving “rapid growth of distributed power generation, distributed storage and advanced demand-response capabilities, which need to be orchestrated and integrated through more networked, transactional power grids.” AI is therefore essential for the unprecedented demands of decarbonization, which will depend on an intelligent, responsive, and flexible computing system able to recognize and predict complex patterns of production and consumption. But there’s a problem. While AI is necessary to curb emissions, AI itself requires vast amounts of energy to fuel the training and machine learning processes that make the model useful. Certain single AI training models have been shown to use the equivalent of 125 New York-Beijing round-trip flights, or the lifetime carbon footprint of five cars.

So, is AI a net positive for energy efficiency and greenhouse gas emissions? Not always, according to a recent report from Semiconductor Engineering. Using AI responsibly and efficiently requires a number of considerations and calculations. Starting with the simple question: does this system actually need AI? While artificial intelligence undeniably has much to offer to the energy industry, it can also be more seductive than strictly necessary in certain contexts. In the words of Semiconductor Engineering, “we can no longer afford to be profligate with our resources; we need to ensure that the benefit outweighs the cost.”

If the system in question would indeed have net benefits from AI, engineers will next have to consider where the energy for the training is sourced, whether workloads are designed efficiently and effectively, calculate and consider embedded emissions, and maximize performance per watt.

If AI is optimized for maximum energy efficiency and trained using clean energy sources, it’s a no-brainer for the energy transition. But making responsible, effective, and climate-conscious AI capable of catalyzing the clean energy revolution will require ‘clear policy incentives,’ but these have not been forthcoming, as AI is still relatively poorly understood and somewhat mistrusted in public spheres. Employing AI to its fullest potential will require a deep understanding of the enormously positive potential benefits it offers, in addition to its potential pitfalls.

Crude Oil