Researchers are pushing tech companies for more transparency about the impact of their AI products on the environment as new reporting shows just how significant it’s likely to be. The increased use of artificial intelligence is already leading to more carbon emissions and higher demand on electricity and water. A growing number of voices are calling on AI developers to share that information with customers.
A search query on OpenAI‘s ChatGPT uses as much as 10 times the power consumption of a standard Google search, while the AI chatbot requires the equivalent of 16 ounces of water to process 10 queries, according to Shaolei Ren, associate professor of electrical and computer engineering at UC Riverside. These levels of consumption cast the industry’s sustainability into question.
As access to water becomes more difficult in some areas of the U.S. and other countries, governments are facing shortages. AI will only make those worse. The specialized computer chips they require are power-intensive and need water to cool them. The increased demand caused by AI and the data centers that power the technology could also slow the transition to green energy and drive up consumer electric bills and make blackouts a more common occurrence.
It’s estimated that global AI-related electricity consumption could rise by 64 percent by 2027, using as much energy as Sweden or the Netherlands. Amazon’s AWS cloud subsidiary recently purchased a nuclear powered data center in Pennsylvania in order to co-locate its growing AI data center next to the nuclear power plant that fuels it, and other vendors are also looking for new power sources to supplement or replace fossil fuels.
But Ren and his colleague Alex de Vries—whose company Digiconomist has a mission of exposing the digital world’s unintended consequences—want to give users more agency in the decision. In a recent L.A. Times article, they told reporter Melody Petersen they are joining a growing number of other experts calling on tech companies to disclose the power and water usage of queries to their customers.
As more and more companies incorporate AI into their products—often without having a clear view of the benefits—the demand for power and water and the drain on natural resources will only worsen. Because AI developers “tend to be secretive about their energy usage and their water consumption,” Ren said, he wants them to be more direct in disclosing the ramifications of using their tools to let them make more informed decisions.