The environmental impact of using Artificial Intelligence : Exploring the Environmental Impact of ChatGPT: A Literature Study and User Perception Analysis

Sammanfattning

One of the most transformative technological advancements in recent years has been Artificial Intelligence (AI). The ability to help and assist with tasks in sectors from finance to healthcare has made AI in high demand. In particular, so-called generative AI has captured significant attention, which means AI systems capable of producing new content such as images, text, code or audio based on input prompts. This study investigates the environmental impact of generative AI with a primary focus on ChatGPT. The research is a combination of a literature review of ChatGPT’s water and energy consumption with a user study investigating user behaviour and perception of the environmental impact of utilising it. The findings indicate that the training and operations of ChatGPT require considerable energy both for training the algorithms and for their operations in use. The estimated electricity consumption of ChatGPT is significant for the maintenance and processing of data. Moreover, water consumption, used primarily for cooling data centres, contributes to the environmental impact of ChatGPT and other AI services by increasing freshwater scarcity in regions already affected. The user perception survey revealed that most participants were not actively thinking of the environmental impact that ChatGPT has while using the model. Many participants recognised the size and advancement of ChatGPT and the energy consumption for it would be similar but were unaware of the water consumption of the model. 25% of the participants indicated they would change their usage behaviour if they had more knowledge about the ecological footprint of ChatGPT and 47.2% thought they might change their behaviour. This study sheds light on the water and energy consumption of AI to contribute to a more environmentally responsible approach to AI development and usage.