Artificial Intelligence (AI) is a double-edged sword. On the one hand, the AI environment presents considerable upsides, on the other it has potential environmental pitfalls. The World Economic Forum makes the case for a four-step strategy to leverage AI responsibly.
The Dual Role of AI Environment
Artificial Intelligence (AI) has the potential to significantly contribute to environmental conservation by lowering greenhouse gas emissions and enhancing the utilization of natural resources. However, it can also inadvertently escalate demand, subsequently amplifying its carbon footprint. A four-pronged approach is proposed to guide the responsible deployment of AI.
The Unprecedented Power and Peril of AI
AI’s capabilities to handle vast amounts of specific data, identify patterns, and derive autonomous decisions offer ground-breaking advancements across numerous fields. AI is being utilized to transform operations across a multitude of sectors, including sales, marketing, customer service, manufacturing, research and development, IT, human resources, and finance. AI is even being integrated into workforces via “cobots” – collaborative robots that operate alongside humans to enhance productivity and ultimately improve product quality.
The burgeoning use and broad impact of AI necessitate a critical examination of its potential environmental ramifications. Could the considerable computational power demanded by AI have a detrimental effect on the environment, or could AI emerge as an extraordinary tool that propels economic growth without environmental degradation?
The Positive Impacts of AI
AI can offer substantial environmental benefits in numerous ways:
Reducing Carbon Emissions
According to a joint study by Microsoft and PwC in 2019, conscientious usage of AI could lead to a 4% (2.4 giga tonnes) reduction in global greenhouse gas emissions by 2030. AI is currently being utilized to streamline industrial and residential energy consumption, thus reducing their respective carbon footprints. Furthermore, AI is being deployed to optimize logistics, predict supply and demand, schedule maintenance, and manage autonomous transportation, leading to direct and indirect decreases in carbon footprints.
A striking example is Google’s use of AI to decrease the energy consumption of its data centers. By utilizing machine learning technology developed by DeepMind and optimization algorithms, Google has improved its data center’s energy efficiency by 35%.
Effective Utilization of Natural Resources
AI is also aiding in predicting the energy yield from renewable sources such as solar, wind, and hydro energy, thus minimizing wastage of these resources. AI assists in conserving water in residential, manufacturing, and agricultural sectors. Predictive AI algorithms have led to innovative agricultural techniques like precision farming, ensuring optimal water usage and harvest of mature crops. These algorithms also contribute to farmland planning, crop and livestock health monitoring, and efficient power-generation planning for power generators and consumers.
An innovative initiative, the California Forest Observatory (CFO), developed by Salo Sciences, Vibrant Planet, and Planet Labs, uses an AI engine to map wildfire behaviors across California. Leveraging LIDAR and satellite data, CFO provides detailed forest structure and fuel load data, which, combined with wind, weather, soil, and vegetation moisture data, helps capture complex wildfire risk drivers.
The Environmental Trade-offs of AI
Despite its potential for remarkable process improvements, AI is not without its drawbacks. A study by the University of Massachusetts Amherst found that training a large AI model would have a carbon footprint 300 times larger than a flight from San Francisco to New York.
AI, even when used with positive intentions, can lead to unintended consequences. By optimizing business operations, AI can inadvertently increase demand and thus carbon footprint.
A Four-Pronged Approach to Environmentally Responsible AI Usage
To responsibly harness the potential of AI, the following four-step strategy is recommended:
- Choose the right use case: Not all optimizations result in significant carbon footprint reductions. AI can address a vast array of challenges, and users need to prioritize those processes that can truly be optimized by AI for sustainability purposes.
- Choose the correct algorithm: The process of training an algorithm significantly impacts its energy consumption. Choosing the right AI algorithm for a given problem or use case can dramatically decrease energy requirements.
- Monitor carbon footprint outcomes: Sustainability should be considered a vital success indicator in any AI project. Companies are beginning to include carbon footprint estimates in their cost/benefit analyses for responsible AI deployment.
- Balance carbon footprint with renewable energy: Offset carbon footprint by powering AI models with renewable energy. Google, acknowledging the high energy consumption of AI in its data centers, pledged to power its data centers with renewable energy and has maintained net-zero carbon emissions since 2017.
Tools like CodeCarbon can assist organizations in implementing these steps. This lightweight, open-source software package, jointly developed by Mila, BCG GAMMA, Haverford College, and Comet, tracks the carbon dioxide (CO2) produced by cloud or local computing resources used in AI experiments and provides insights into the CO2 impact.
AI: A Powerful Ally in Combating Climate Change
AI has rapidly emerged as one of the most influential tools at humanity’s disposal. As with any revolutionary technology, it carries the potential for both positive and negative impacts. When used judiciously, AI could prove invaluable in our fight against the escalating climate crisis.
See the original report on the World Economic Forum’s website.
