Climate Change in the Age of AI

By: Anita Voskovykh

Artificial intelligence (AI) has grown exponentially in both its development and use, and in many ways has become inescapable. AI is technology that collects, processes, and uses data to simulate human intelligence processes via technological machines, most commonly computers. With climate change considered to be “the defining crisis of our time,” the continuous advancements and investments in AI have the potential to become a major tool in the fight against climate change and may ultimately benefit society as a whole if implemented carefully. However, these advancements come with inherent risks, including environmental impacts as a major concern given the energy intensive process that these developing AI models demand.

 With its growing complexity, the power and energy consumption required to train emerging AI models increases as well. The training involved in AI models alone consumes thousands of megawatt hours of electricity and emits a significant amount of carbon into the air. Moreover, this environmental toll will likely continue as demand for more advanced AI technology grows. A large carbon footprint attaches to the development, maintenance, and disposal of AI given its reliance on energy-intensive processes. For example, a Goldman Sachs report revealed that a request made on ChatGPT consumes 10 times more energy than a Google search. Another study from the University of Massachusetts found that “training an “off-the shelf” AI using a single high-performance graphics card has the same carbon footprint as a one-person round trip flight between New York and San Francisco.” The overall production, transport, maintenance, and disposal of hardware components used in facilitating AI, such as servers and data centers, require additional energy consumption and further the tremendous strain on the environment. 

The AI industry is trending towards prioritizing accuracy over efficiency. This, in turn, requires more data to be used for trial-and-error training. Training AI models requires significant exposure to great amounts of data. These data are referred to as “training data” and are designed to run through neural networks that assign parameters on what the desired outcome is. The AI relies on this massive amount of data due to the trial-and-error nature of this process which is repeated over and over again until the AI finally produces the intended result. These training mechanics contribute heavily to the consumption of resources that cause AI to become environmentally problematic. However, this is also one the most essential components in the overall advancement of AI. Without it, AI models will likely not produce the results necessary to be valuable to society. 

 Despite the environmental harms associated with AI, it is also important to recognize its potential advantages associated. Much of the discussion surrounding the implementation of AI focuses on how it can improve sustainability efforts and tackle climate-change related issues. The emerging AI models bear promising climate-related solutions. Some of the most significant of these recent models include monitoring weather and new patterns of pollution, managing current agricultural systems, tracking melting icebergs and air emissions, mapping deforestation, and preparing for natural disaster and the aftermath of it by identifying buildings damaged by natural disasters. While AI use for the mitigation of climate-change related issues is promising, the AI industry must consider the immense impact it has on the environment and implement methods to find a balance. Without this, any potential climate solutions will be outweighed by the disproportionate disruptions to the environment AI causes.  

 AI has indeed revolutionized our world and will likely continue on this path. Regulations on the ethics and monitoring of AI use are likely to become more frequent and common for this emerging technology. Developers, those with expertise in AI, and lawmakers should establish principles and ethical limitations on the use of AI with the goal of avoiding further climate damage in mind. Recent government efforts on AI legislation have already begun making their way through the US. At least 45 states in the 2024 legislative session have introduced AI bills, and 31 states have adopted resolutions or enacted legislation. Further, Hawaii has required the University of Hawaii to implement a program to develop a wildfire forecast system to forecast the risk of wildfire state using AI. With all of the potential benefits associated with AI also comes an immense responsibility for lawmakers, tech-developers, and even society to ensure that it is not at the environment’s expense. Legislative development is crucial for emerging AI models, as the potential for improvement also carries the risk of undermining any progress that has already been made.