Professor Manel Poch, from the University of Girona, points out that the ease of opening the faucet and that drinking water of the best quality is available at low and subsidized prices -less than one cent per liter- makes us forget the infrastructure that this requires and its cost. Only the current drought once again draws our attention to this precious resource. The same happens with the costs of using the Internet and associated computer applications, of which we think are almost free and do not have an effect on the environment.
Technologies based on Artificial Intelligence (AI) have tremendous potential to support positive climate action. These technologies support the green transition, from digital twin technology that attempts to model a building, a city, or the Earth to algorithms for making data centers have optimal energy consumption using renewable energy. However, AI-based systems They also raise sustainability issues. related to the natural resources they consume, such as electricity and water, and the carbon emissions they produce. The rise of deep learning (Deep Learning) and the applications of large language models (Large Language Models) -such as the already famous ChatGPT- have also increased drastically. the amount of computation needed to train them and then to keep them running. Some of them serve hundreds of millions of users each. For example, the electricity consumption of ChatGPT in January 2023 was between 1,168,200 KWh and 23,364,000 KWh, which is equivalent to the consumption of some 175,000 citizens in the same period in the US. To these figures should be added what each user consumed when interacting with these applications.
Related news
Other data that can help us understand the environmental impact is that more than 80% of the systems that are trained using Deep Learning never reach the market. That is, wasted energy. Like supercomputing centers, big tech is exploring solutions to environmental impacts, such as intensive use of renewable energy sources to feed the development and exploitation of these systems.
We are not aware of the relationship between the use of digital technologies and their environmental effect. Being more digital today does not imply that the extensive use of these technologies has a neutral environmental impact and, therefore, that as a society we are greener. As research and widespread use increase of intelligent systems, governments, legislators and citizens must understand the environmental impacts of AI to make decisions based on the already well-known evidence. This raises a question for us as a society: how much is it worth the environmental footprint caused by the use of smart applications?