Since the launch of ChatGPT, 200 million people use the platform daily. From college students to business professionals, ChatGPT has revolutionized the search for information. But if you are someone like me, interested in sustainable AI or using AI in a way that minimizes its environmental impact, ChatGPT is not the “Green AI” tool you may have believed it to be. For example, Forbes reports that “a single ChatGPT conversation uses about 50 centiliters of water, equivalent to one plastic bottle” and “the 200 million daily users of ChatGPT are consuming as much electricity as 180,000 households do.” These facts made me ponder a paradox in the accounting and finance world related to AI and ESG reporting; as AI becomes necessary for ESG reporting, it is also a contributor to an organization’s greenhouse gas emissions.
How does AI generate greenhouse gases? AI systems collect and process large volumes of data, requiring significant computing power and a power source (i.e. electricity) to run it. These AI systems live in data centers, which account for 2% of global electricity demand, according to Forbes. Data-intensive applications like AI are predicted to increase 80% worldwide from 2022 to 2026, according to the International Energy Agency. These centers also produce electronic waste and use water during construction and during operations to cool parts at a time when access to clean water is a problem for many worldwide.
Why can’t ESG reporting be done without AI? Answering that question warrants a look at what is involved in the ESG reporting process itself. For example, KPMG outlined the steps a factory would need to take to track its carbon footprint for reporting purposes. The data needed for this exercise must be extracted from energy bills, transport logs, fuel receipts, and production records. Different departments across the business supply the data in different formats, without aggregating it or confirming its accuracy. Enter generative AI, which can analyze vast amounts of data, identify patterns, and provide insights into ESG performance. It can also prepare reports and clarify questions for users across the business who all have access to one data dashboard, serving as a single source of truth across the organization. Without AI, there is no single source of truth and no way to verify the accuracy of data used for ESG reporting.
The good news is that companies like Microsoft are actively working on ways to reduce the environmental impacts of AI. Part of that strategy includes improving how AI-based systems use renewable energy and reducing the cost of carbon capture. Finding ways to make AI systems run more efficiently (using less processor time and memory) is another key objective. Globally, this issue is also on the radar of the UN Energy Program which is spearheading discussion among countries about the ethical use of AI and formulating strategies to mitigate AI’s environmental impact.
As accounting and finance professionals are asked to help their organizations devise strategies for minimizing greenhouse gas emissions, as well as capture, measure, and report on ESG performance, the AI-systems they use for these activities must also be included in ESG reporting as greenhouse gas emitters. This paradox mirrors others when it comes to sustainability related measures. For instance, if I buy an electric car and I don’t have solar panels, charging my car will just result in more household electricity usage.
There are no easy answers when it comes to sustainability. Accounting and finance professionals sit where these issues converge so awareness of the paradoxes can be helpful in guiding organizations further along the sustainability continuum.