AI is on the lips of climate-policy negotiators gathered for the United Nation’s COP28 conference in Dubai, and for good reason — it presents a high-risk but potentially high-reward scenario.
The upside: AI has the potential to supercharge efforts to find real climate solutions. For example, scientists can send AI-powered robots to collect data in the Arctic and other challenging environs, and the technology can also be used to improve forecasting for extreme weather and climate-related disasters. On an even more basic level, it can be used to maximize the efficiency of all kinds of systems and reduce their carbon footprint.
But there’s a big catch: AI is an energy-guzzler. One analysis found that AI systems worldwide could consume 85 to 134 terawatt-hours per year — equivalent to the electricity diet of Argentina or the Netherlands. That’d be good for half a percent of the world’s energy consumption. (This analysis is based on the sale of popular servers from US chipmaker NVIDIA, used by much of the AI market.)
At COP28, government and industry leaders made bold announcements. Boston Consulting Group said AI could reduce greenhouse-gas emissions by 5-10% by 2023. Meanwhile, the UN announced a deal with Microsoft to use AI to track countries' carbon-reduction promises.
Is the risk worth the reward? “Whether you like it or not,” says Shari Friedman, managing director for climate and sustainability at Eurasia Group, “AI is here to stay, so the job of humans will be to use it for the best purpose possible and maximize clean energy on the back end.”