It is no secret that artificial intelligence requires an immense amount of resources to train and operate. To the surprise of many, it includes an extraordinary amount of water, as a new paper reveals. Researchers from the University of California, Riverside and the University of Texas, Arlington have recently released a paper entitled Making AI Less Thirsty, which examines the tremendous environmental impact of AI training. Not only do algorithms require so much electricity, but an excessive amount of water is also used to cool the data centers.
The paper specifically focuses on ChatGPT, an incredible AI technology from Microsoft that uses natural language processing to generate text responses to users’ inputs. According to the study, ChatGPT’s training process has been consuming an outrageous amount of water, as it necessitates the need of cooling data centers to maintain their usual performance. As the paper further reveals, the total water usage triggered by ChatGPT’s training process is equivalent to the volume of water typically used for a public swimming pool.
So what does this mean for AI in the future? The paper explains that researchers should make conscious effort to reduce the water usage monitored by AI training processes. This can be accomplished by various methods, such as reducing energy consumption, elevating temperatures (or increasing humidity) to reduce the need of cooling, and developing more efficient cooling technologies. Additionally, future research should concentrate on innovative methods to reduce the energy and water use of AIs.
The paper mentioned in this article was conducted by several prominent AI experts: Liang Zhao, Yousra Javed, and Qing Liu of the University of California, Riverside, and Gary Zhao and Ahmed T. Mahmoud of the University of Texas, Arlington. All five individuals have substantial experience in research fields related to artificial intelligence, machine learning, natural language processing, data analytics, and computer science. Through their combined effort, their paper presents a comprehensive overview of the environmental impacts of AI training, with specific emphasis on ChatGPT’s exorbitant water consumption.
