COP31 highlights responsible AI for climate

At the major annual United Nations climate conference in November in Turkey, COP31, the organizers aim to promote a "responsible AI for climate." Siddharth Singh, an AI specialist at the International Energy Agency (IEA), explains to AFP that data centers represent a real challenge, with concerns about their growth's implications for energy systems, the energy sector, and the climate being legitimate. However, he adds that AI "is also a tool and represents an opportunity in that sense. It all depends on how it is used."

AI applications are less energy-intensive

According to several experts, most of these applications are actually much less energy-intensive than the models used for ChatGPT, Claude, or Copilot. AI is already being used to optimize electrical grids, for example, to better manage the intermittency of renewable energy. Thanks to highly localized forecasts of cloud cover, sunshine, or wind, the technology allows for more precise anticipation of renewable electricity production and thus limits periods when wind turbines or solar panels must stop because the grid cannot absorb all the available energy.

AI could save significant energy by 2035

According to the IEA, if it were deployed on a large scale in transport, industry, and buildings, AI could save 13.5 exajoules of energy by 2035, equivalent to the annual energy consumption of Indonesia. At the individual level, automatically programming a washing machine when there is an excess of solar electricity available thanks to AI can reduce its carbon footprint, explains Siddharth Singh. The environmental organization WattTime is already helping companies program the charging of electric vehicles, phones, or other devices during periods when electricity comes from sources that emit less greenhouse gases.

AI improves weather forecasting

"Forecasting is one of the strengths of AI," says Gavin McCormick, co-founder of WattTime. AI has truly revolutionized weather forecasting because it is now possible to predict the weather much faster, with a very high level of accuracy and performance, states Florian Pappenberger, director general of the European Centre for Medium-Range Weather Forecasts (ECMWF). AI-based forecasts still rely on classical physical models, but they allow the ECMWF to accomplish in two years what would have previously required ten years of work, he explains to AFP. They are also much less costly than traditional approaches, which allows for more forecasts to be produced.

AI enhances climate warning systems

AI can also be used to improve warning systems; Google claims to be able to anticipate river floods up to seven days in advance in more than 150 countries. Rather than replacing humans or existing methods, "it's as if we were given superpowers to do even better what we already did," says Florian Pappenberger. Combined with satellites, AI can more easily detect greenhouse gas emissions from oil wells, mines, or landfills. Climate TRACE, an organization co-founded by Gavin McCormick, has thus been able to establish a detailed inventory of pollutants emanating from sites around the world. "Thanks to AI, Climate TRACE has taken twenty years ahead," claims McCormick. "We are able to provide emission inventories to all cities, which can use them to develop climate action plans. Without AI, this simply would not have been possible."

AI has dual impacts on climate

The fossil fuel sector has long used AI to improve its production by analyzing vast volumes of geological data and locating underground resources more effectively. "From this perspective, AI can also have a negative impact on the climate," notes Maria Joao Sousa, director of the organization Climate Change AI, which promotes the use of artificial intelligence in climate action. For her, "the technology alone will never be the solution."