Generative AI and Climate Change are in Conflict Studies

By 2025, AI and climate change, the two biggest societal disruptors we face, will collide.

The summer of 2024 broke the record for the hottest day on Earth since data collection began, sparking media coverage and public debate. This was also the year that both Microsoft and Google, two leading technology companies investing heavily in AI research and development, missed their climate targets. While this has once again made headlines and fueled outrage, the effects of AI on the environment are far from common knowledge.

In fact, the current AI paradigm of “bigger is better”—symbolized by tech companies’ pursuit of large, powerful language models presented as the solution to all problems—comes with significant environmental costs. This ranges from generating large amounts of energy to power data centers using tools like ChatGPT and Midjourney to millions of gallons of fresh water being pumped into these data centers to ensure they don’t overheat and tons of rare earth metals. necessary to build the hardware they contain.

Data centers already use 2 percent of the world’s electricity. In countries like Ireland, that figure rises to one-fifth of the electricity produced, which has prompted the Irish government to announce the effective suspension of new data centers until 2028. Although most of the energy used to power the data centers is officially “carbon-neutral,” this depends on methods such as renewable energy credits, which technically reduce the emissions caused by the production of this gas, but do not change the way it is produced.

Areas like Data Center Alley’ in Virginia are largely powered by non-renewable energy sources like natural gas, and energy providers are slowing the phase-out of coal-fired power plants to keep up with the growing demands of technologies like AI. Data centers are draining large amounts of fresh water from scarce groundwater resources, pitting local communities against data center providers in places ranging from Arizona to Spain. In Taiwan, the government has chosen to allocate precious water resources to chip factories to keep up with growing demand instead of letting local farmers use them to irrigate their crops during the country’s worst drought in more than a century.

My recent research shows that going from old general AI models—trained to do a single task like answering questions—to new generative models can use up to 30 times more power just to answer the same set of questions. Tech companies that are increasingly adding artificial intelligence models to everything from search engines to word processing software also don’t reveal the carbon costs of these changes—we still don’t know how much energy is being used during a chat with ChatGPT or in production. image with Google Gemini.

Much of the talk from Big Tech about the environmental impacts of AI has followed two paths: Either it’s not really a problem (according to Bill Gates), or an energy boost will appear and magically fix things (according to Sam Altman). What we really need is more transparency about the environmental impacts of AI, in the form of voluntary initiatives like the AI ​​Energy Star project I lead, which can help users compare the energy efficiency of AI models to make informed decisions. I predict that by 2025, voluntary programs like this will begin to be implemented by law, from national governments to governmental organizations such as the United Nations. By 2025, with more research, public awareness, and regulations, we will finally begin to grasp the environmental footprint of AI and take the necessary steps to reduce it.