Artificial intelligence: The use of electricity in AI-related activities could increase, surpassing the annual electricity demand of countries like the Netherlands, Argentina, and Sweden by 2027.
New research suggests that by 2027, activities involving electricity-powered Artificial Intelligence (AI) could surge. This could lead to a greater demand for electricity used by AI than what countries like the Netherlands, Argentina, and Sweden require annually.
Why is this happening? A claim is made in a commentary published in the Joule journal. The research states that adopting AI on a broad scale may require substantial energy consumption because technologies like GPT-3, which generates generated content based on trained data, have been rapidly advancing since 2022.
According to Dr. Freek, a researcher from Vrije Universiteit Amsterdam, the training of these models is an energy-intensive process, as it involves feeding a large amount of data into the models.
Hugging Face, a New York-based AI company reported that it consumed 433 megawatt-hours (MWh) of electricity to train its text content generative AI tool, which is enough to power around 40 average American homes for a year. Furthermore, these AI tools consume energy each time they generate text or images based on input signals, contributing significantly to energy consumption.
Dr Freek’s analysis revealed that running GPT-3 could potentially consume 564 megawatts of electricity per day. He also estimated that if Google, which currently processes up to 9 billion searches per day, were to empower its search engines with AI, it would require approximately 30 TWh of electricity per year, roughly equivalent to Ireland’s annual electricity consumption. Given the increasing demand for AI services, it is highly likely that energy consumption related to AI will see significant growth in the coming years.