The environmental impacts of AI

The rapid rise of AI use

It’s hard to read, listen to or watch any news outlet without seeing something about artificial intelligence (AI). The AI market is projected to reach $407 billion by 2027 with an annual growth rate of 37.3% from 2023 to 2030 [1]. Organizations continue to find innovative ways to leverage AI for increased productivity and efficiency. At the same time, the world is struggling to figure out if and how to regulate these powerful tools. Like many things in life, the key is finding the right balance, and there are several trade-offs to keep in mind.

Trade-offs for the environment

With increased data usage also comes an increased carbon footprint. Not surprisingly, the data sets required to train and run AI models are large and often complex, leading to higher energy consumption. According to an MIT study, the cloud now has a larger carbon footprint than the entire airline industry and training a single AI model can emit more than 626,000 pounds of carbon dioxide equivalent [2]. As many organizations are monitoring their carbon footprints closely, this adds another layer of consideration to decisions made around data centers, machine learning and energy use.

At the same time, AI has an important role in solving complex environmental problems. Its ability to quickly aggregate data from multiple sources has been used to help inform real-time analysis and monitoring of emissions, air quality, weather disasters and environmental impacts. In a recent study by BCG, 87% of climate and AI leaders found AI to be a helpful tool in the fight against climate change, with reducing and measuring emissions cited as the top drivers for creating business value [3].

More work and research need to be done to strike the right balance between using these tools and minimizing their environmental impact. This starts with developing awareness and understanding of the tools and their side effects. Organizations will need to calculate the impact of their AI and machine learning use, examine where data is stored (including the energy sources for the storage) and design data models as efficiently as possible [2].

Final thoughts

It’s difficult to say whether the adverse energy effects outweigh the positive efficiencies gained through AI tools. As our daily use of these tools continues to grow through ChatGPT and machine learning, it is important to be mindful of these tradeoffs and continue to conduct due diligence to understand the full impacts of our choices.