A new framework for understanding and shaping the effects of AI on greenhouse gas emissions.
Will the increasingly widespread use of artificial intelligence (AI) aid or hinder the world's fight against climate change? A team of AI, climate change, and public policy experts present a framework for understanding the complex and multifaceted relationship of AI with greenhouse gas emissions in a paper published this week in Nature Climate Change and suggest ways to better align AI with climate change goals.
"AI affects the climate in many ways, both positive and negative, and most of these effects are poorly quantifiable," said co-author David Rolnick, Assistant Professor of Computer Science at McGill University and Core Academic Member of Mila — Quebec AI Institute. "For example, while AI is being used to track and reduce deforestation, AI-based advertising systems are likely to exacerbate climate change by increasing the amount that people buy."
The paper categorizes the effects of AI on greenhouse gas emissions into three groups: 1) Impacts from the computational energy and hardware used to develop, train, and run AI algorithms; 2) immediate impacts caused by AI applications, such as optimizing energy use in buildings (which reduces emissions) or accelerating fossil fuel exploration (which increases emissions); and 3) system-level impacts caused by the ways in which AI applications affect behavior patterns and society more broadly, such as through advertising systems and self-driving cars.
" Climate change should be a key consideration when developing and evaluating AI technologies," said Lynn Kaack, Assistant Professor of Computer Science and Public Policy at the Hertie School of Governance and report lead author. "We discovered that the impacts that are easiest to quantify are not always the ones with the greatest impact. As a result, it is critical to assess AI's impact on the climate holistically."
The effects of AI on greenhouse gas emissions are a matter of choice.
The authors stress the ability of researchers, engineers, and policymakers to shape the effects of AI, writing that its "... ultimate effect on the climate is far from predetermined, and societal decisions will play a large role in shaping its overall impacts." The paper, for example, notes that AI-enabled autonomous vehicle technologies can help lower emissions if they are designed to facilitate public transportation, but they can increase emissions if they are used in personal cars and cause people to drive more.
The researchers also point out that machine learning expertise is frequently concentrated among a small number of actors. This raises potential challenges for machine learning governance and implementation in the context of climate change, as it may create or widen the digital divide, or shift power from public to large private entities based on who controls relevant data or intellectual capital.
"The implicit choices we make as technologists can have a big impact," Prof. Rolnick said. "Ultimately, AI for Good should be about shaping all AI applications to achieve the impact we want to see, not just adding beneficial applications on top of business as usual."
Source: Materials provided by McGill University.
Reference: DOI: 10.1038/s41558-022-01377-7