How can policy help regarding AI’s environmental impact?
Currently, governments are looking for ways to deal with the rapid growth of AI. What those data centers might look like is also up to the regulations our governments create, and what we choose to do with it. Here are some ideas that might shape the creation of new policies regarding AI and data centers’ environmental effects, and some ways policy might improve the environmental impact of AI’s demand.
One of the biggest obstacles in measuring the environmental impact of AI is the limited data and information available from the developers of the technology. There is considerable uncertainty regarding the amount of energy and water required for data centers, as data center companies and utilities rarely provide public reports on this information. More transparent and detailed reporting is needed at both the state and federal levels to ensure that we do not overbuild the electricity system and to protect other ratepayers from paying any excess or stranded costs. There are many steps to training and using an AI model. Breaking down the resource usage of each of these steps could help us understand how to mitigate their impact 1. Also, since AI has many indirect environmental impacts, like the impact of building the chips, requiring companies to have more detailed reports about their use of resources like energy and water, and the emissions they generate, could lead to policies more effective at preventing or reducing negative environmental and public health impacts 2.