AI’s energy consumption is a serious concern. We touched on it in several places on the Will AI Automate Most White Collar Jobs thread.
Training a single large language model consumes as much electricity as hundreds of homes use in a year, and data centers that power AI are expanding rapidly with much of the electricity still coming from fossil fuels. Water usage is another problem especially here in the desert where many large data centers are located due to the dearth of natural weather disasters. Training AI models consumes millions of gallons of water in equipment cooling systems. Hardware and electronic waste costs include mining rare earth metals, increasing manufacturing emissions, and short hardware life cycles. As AI demand rises, the turnover of high-performance computing hardware is rising sharply, too. Carbon emissions are problematic as well when data centers are powered by coal-heavy grids, but we’re making headway in using water, wind, solar, and geothermal resources.
There is a balance. AI has the potential to accelerate climate solutions and bring efficiency gains across industries, but the net impact will depend on how responsibly AI is developed. We’ll have to see. This article from MIT acknowledges the problem but also outlines some solutions which include limiting the amount of power available, rethinking how models are trained, and making software AI- and carbon-aware.
While the MIT Lincoln Laboratory Supercomputing Center continues to use these energy-efficient AI initiatives, there’s more work to be done, Gadepally said, including working with the broader community to collect data, build benchmarks, and rethink the current mindset that casts bigger AI models and data as better. “We need to think more about how we can get to the same answer but add a bit of intelligence to make AI processing more energy efficient,” he said.