Artificial intelligence (AI) technology could have a more significant environmental impact than previously anticipated, according to a recent study published in the journal Joule. The research suggests that AI’s energy consumption could eventually reach levels comparable to an entire country, such as Sweden. This revelation poses a significant challenge to global efforts aimed at reducing carbon emissions.
The study, conducted by Alex de Vries, a PhD candidate at the VU Amsterdam School of Business and Economics, highlights the rapid advancement of AI technology. De Vries emphasizes that as AI progresses, large language models (LLMs) like ChatGPT require substantial datasets for training. This results in a significant waste of power and hampers efforts to curb energy consumption.
While previous studies have focused on the energy consumed during AI training, De Vries underscores the need to consider the full life cycle of AI. He points out that the inference phase, where trained AI models generate information based on new inputs, accounts for a substantial portion of energy consumption. In fact, Google reported that 60% of its AI-related energy consumption between 2019 and 2021 originated from the inference phase.
- 5 AI-powered presentation tools to help you at work
- 90s AI Yearbook: How to generate your own retro-themed images
With energy production being responsible for more than three-quarters of global greenhouse gas emissions, the increasing energy footprint of AI poses a significant concern. Pumping greenhouse gases into the atmosphere contributes to climate change, and urgent action is required to mitigate its effects. The study estimates that by 2027, newly manufactured AI devices alone could consume as much electricity as an entire country, such as the Netherlands, Sweden, or Argentina.
As AI products become more prevalent and adopted by numerous companies, the demand for AI chips is surging. Nvidia, for instance, reported a 141% increase in its data center segment, primarily driven by the growing demand for AI products. Companies like Alphabet’s Google, with its integration of generative AI into every search, could significantly contribute to the rise in AI’s energy consumption.
Despite the potential environmental impacts of AI, De Vries believes that as the technology evolves, so will the supporting infrastructure. He emphasizes the importance of AI developers being mindful of when and how they utilize the technology. While AI can bring many benefits, it is not a magic fix for all problems and should be used carefully.