Environmental impact of artificial intelligence: How much does each query pollute?

Although AI facilitates everyday tasks, recent research highlights that the energy efficiency of these systems is a growing problem.

• Recent studies reveal that a single query can emit much more CO₂ than a conventional search.

Generative artificial intelligence (AI) tools have become essential in many people’s daily lives. Although they are seen as quick and effective solutions, research cited by CNN points out that, behind every word generated by these machines, there is a high environmental cost. How generative AI models work involves breaking down the instructions we give into a set of numbers called token IDs. These tokens are sent to data centers, where computers perform complex calculations to generate the requested responses. The electricity used comes from power plants powered by coal or natural gas, sources that contribute to gas emissions. According to the Electric Power Research Institute, processing a query on an AI model can consume up to ten times more energy than performing a conventional Google search. In a study conducted in Germany, researchers evaluated 14 AI systems based on large language models (LLMs). They were asked open-ended and multiple-choice questions to measure the environmental impact of each. Complex questions generated up to six times more CO₂ emissions than those with simpler answers. The most advanced models, which have greater reasoning capabilities, produced up to 50 times more emissions than the smaller models when answering the same query.

Continue reading news in Entrepreneur