Generative AI became widely recognized during my sophomore year of high school 2022-2023. In the wake of this new technology, I suffered through countless lectures about the ethical dilemmas surrounding the use of AI in schools. In those conversations, the environmental impact of AI did not come up. As my peers and teachers were finding ways to utilize the capabilities of AI to make life easier, the cost of convenience never crossed their minds.
Much like driving to school is easier than biking, but emits more pollution, the seemingly limitless power behind AI is dominating our power grid. It’s estimated one Chat-GPT query (Basically a question for a computer) uses 2.9 watts of electricity compared to 0.3 for an average Google search. The bulk of AI’s energy consumption comes from training done before it is even released to the public. This process consists of feeding the machine data curated to allow the AI to produce accurate and specific results to questions it may be asked. This becomes an issue when AI models need to be retrained frequently, making sure they have access to up-to-date information.
But the impacts of AI don’t only come from the usage of AI, but the physical space it takes up. It’s easy to assume that ChatGPT and other AI models exist only in space or some imaginary ‘cloud’. But in reality, AI needs physical data centers. The infrastructure for AI and the ‘cloud’, accounts for more greenhouse gas emissions than all commercial flights globally in a year.
Water is also exploited within these giant data centers. Research suggests that the amount of water needed to cool machines located at Microsoft’s ChatGPT-3 is around 700,000 liters. That fills up about 583333.33 40oz Stanley cups.
With the rapid rise of generative AI technologies, it may seem almost impossible to halt the damage these programs are doing to the environment, but it is possible. One of the first steps to lower AI’s footprint is to demand transparency from the large tech companies that control the AI market. It’s also important to recognize and promote the responsible use of AI. This may look like holding companies accountable for their footprint, using and training AI selectively as to not exploit the resources behind it, or opting for more eco-friendly search engines such as google.
Because AI is so new, there are no precedents for how to navigate the world of AI chatbots. It’s imperative that issues of plagiarism to deflecting workloads, and the heavy impacts on on energy consumption, greenhouse gas emissions, and water use are addressed. Humans and AI can live together, but we must do so responsibly.
