Article: Should AI Supersede our Attitude Towards Nuclear Power? by Dr. Timothy Smith

Photo Source: Wikimedia Commons
The staggering amount of energy needed to run artificial intelligence, especially in the age of large language models, or LLMs, such as Open AI's ChatGPT and Microsoft's CoPilot, may quietly reverse the decline of nuclear power in the United States. A PhD candidate at the Vrije University School of Business and Economics in Amsterdam and the founder of Digiconomist Alex de Vries suggests AI will significantly increase the need for electricity to support data centers worldwide. Datacenters consist of massive clusters of computers that provide search services for companies such as Google's parent company, Alphabet, Microsoft, and others. Alex reports in an article in the research journal Joule that LLMs consume significant electricity during their training phase and continuously after in the inference phase. During the training phase, the model gets built by consuming billions of pages of text that teach the model how to predict what word will follow another based on the context of the question. During the inference phase, the model answers questions and produces reports, summaries, translations, and more content.
Â
Based on his research, Alex notes that ChatGPT consumes 564 megawatts of power per day to run. (cell.com/joule) To put that in perspective, depending on the climate, a medium-sized city of 100,000-500,000 people consumes an average of 500 megawatts of energy per day to function. Much more significantly, if Google used LLM-powered AI in all its services, from maps to searches, they would need 80 gigawatts of power per day. That level of consumption falls slightly short in the entire country of Ireland and more than in hundreds of other countries, including Nigeria, Ecuador, and Slovakia! Such energy consumption creates competition with residents and other businesses and strains the energy infrastructure. In a surprising move in the fall of 2024, Microsoft announced plans to reopen the shuttered nuclear power plant at Three Mile Island on the Susquehanna River in Harrisburg, Pennsylvania, to exclusively power Microsoft's AI ambitions.
Â
Three Mile Island remains the site of the worst nuclear disaster in US history. On March 28, 1979, Unit 2 of the nuclear power station experienced a partial meltdown that released radioactive gases and iodine into the environment. The partial meltdown occurred due to mechanical failure in the cooling system and human error due to a lack of emergency preparedness. The incident stunned the American people and began an anti-nuclear energy attitude both in the US and globally. Later, on April 26, 1986, one of the reactors at the Chornobyl Nuclear Power Plant in Ukraine exploded, releasing massive amounts of radiation into the atmosphere, killing over 30 workers and firefighters. The radiation directly impacted over 8 million people in Ukraine, Poland, Belarus, and other countries and caused thyroid cancer in over 4,000 people. (bbc.com) The explosion has left parts of Chornobyl unsafe for habitation for the next 10,000 years due to radiation contamination.
Â
The unprecedented expansion of AI and the arms race to build more extensive and powerful models has catalyzed massive growth in the tech industry. Nvidia, the maker of the leading chips needed to build LLMs, went from a modest-sized company in 2017 to the most valuable company in the world, surpassing Apple and valued at $3.4 trillion. (barrons.com) This staggering expansion of AI usage will only continue to draw on limited energy resources and challenge regulators to decide if the expansion of nuclear power to feed the AI hunger for electricity warrants reversing the trend to less nuclear power in the US and increasing the potential risk of another nuclear disaster.

Dr. Smith’s career in scientific and information research spans the areas of bioinformatics, artificial intelligence, toxicology, and chemistry. He has published a number of peer-reviewed scientific papers. He has worked over the past seventeen years developing advanced analytics, machine learning, and knowledge management tools to enable research and support high-level decision making. Tim completed his Ph.D. in Toxicology at Cornell University and a Bachelor of Science in chemistry from the University of Washington.
You can buy his book on Amazon in paperback and in kindle format here.

Comments