As more and more people turn to AI-driven chat applications like ChatGPT, the environmental implications of the technology are coming to light. A study conducted by researchers at the University of California, Riverside and the University of Texas, Arlington focused on the water footprint of OpenAI’s popular AI models GPT-3 and GPT-4. According to the study, Microsoft’s training process for GPT-3 used around 700,000 liters, or the equivalent of 185,000 gallons – of fresh water. This is the same amount of water needed to produce 370 BMW cars or 320 Tesla vehicles.
Based on these numbers, ChatGPT requires 500ml, or the equivalent of a 16.9oz., water bottle for every 20 to 50 questions answered. Though this does not seem like much water, when combined with the sheer size of the platform’s user base, the total water footprint for ChatGPT is still quite large. Microsoft and OpenAI have not made any statements in regards to these estimates.
One other AI model studied was Google’s LaMDA. The water consumption used in its training process is difficult to estimate, as Google did not provide comment on the accurate figures. However, it is estimated to be somewhere in the order of million liters, as detailed in a report published in November 2022.
The authors of the aforementioned study strongly emphasize the need to account for water usage alongside carbon usage when assessing true sustainability in AI. Although transparency remains a rarity within this context, increased research and understanding of the implications of AI technology can only lead to better practices down the road.
Google is the world’s leading technology company that has long been known for its innovative approach to creating products and services. Google’s success has been attributed to its focus on innovation, creativity and its ability to continuously deliver new and useful services to its customers.
OpenAI is a research laboratory focused on developing artificial intelligence models. Founded in 2015 by a consortium of investors, OpenAI focuses on creating agents that can solve complex, real-world problems and tasks through the use of AI models such as GPT-3, GPT-4 and LaMDA. Their mission is to develop AI that is beneficial to humanity – this includes making sure the technology is energy efficient and has a minimal environmental impact.
