Volkswacht Bodensee - Behind AI's carbon footprint, a potential climate tool

The issue will be in the spotlight at the COP31 climate summit in Turkey next month, with the hosts saying they will push for "climate-responsible AI".

But the discussion over the technology's environmental impact -- along with fears that it could pose an existential threat to humanity -- has overshadowed its potential as a tool to combat climate change.

AI is being deployed in a broad range of applications, from improving weather forecasts to monitoring industrial emissions and optimising renewable energy.

Many of these applications are less energy-hungry than general-purpose language models such as OpenAI's ChatGPT or Anthropic's Claude, according to experts.

"Data centres have come in as a challenge... There are genuine concerns about what all of this could mean for the energy system, for the energy sector, for climate," Siddharth Singh, an AI expert at the International Energy Agency (IEA), told AFP.

"At the same time, because (AI) is a tool, it's an opportunity. It really comes down to how it is deployed," Singh said.

- Renewables efficiency -

AI is already being used to make electricity grids more efficient in complex energy systems, helping operators manage the variability of solar and wind power.

The technology can produce localised forecasts of cloud cover, sunlight and wind, allowing companies to predict renewable output more accurately.

Singh said it can help limit the amount of "curtailment" -- the practice of reducing certain types of electricity because the grid cannot handle it.

If deployed across the economy, from transport to industry and buildings, the technology could save 13.5 exajoules of energy by 2035, equivalent to Indonesia's annual energy demand, according to the IEA.

Emissions reductions can be made, for example, by using AI to schedule a washing machine to run when solar electricity is abundant, Singh said.

WattTime, an environmental tech nonprofit, helps companies time their electric vehicles, phones or other devices to periods when the energy is coming from cleaner power sources.

"One of the things AI is very good at is forecasting," WattTime co-founder Gavin McCormick said.

- Weather forecasting revolution -

AI is transforming weather forecasting by producing high-quality projections much faster, said Florian Pappenberger, director-general of the European Centre for Medium-Range Weather Forecasts (ECMWF).

"AI has really revolutionised weather forecasting in the sense that you can actually now predict weather much faster, with a really high skill and with a high performance," Pappenberger told AFP.

He said AI forecasting is still grounded in physics-based models, while allowing the ECMWF to achieve in two years what would traditionally have taken 10 years to accomplish.

It is also much cheaper than traditional methods, enabling the ECMWF to produce more forecasts.

The technology can also improve early warning systems.

A Google system can forecast river floods up to seven days in advance in over 150 countries, according to the US tech giant.

Rather than replacing humans, Pappenberger said, "I think (AI) is more like giving us superpowers to be even better or do things even better than we ever did before."

- Spotting emissions -

Satellites and AI are also used to spot greenhouse gas emissions from individual facilities such as oil and gas sites, mines or landfills.

An initiative co-founded by McCormick, Climate TRACE, harnesses the technology to track emissions around the planet and build an inventory of the pollutants.

"Climate TRACE is 20 years more advanced because of AI," McCormick said.

"We have been able to provide emissions inventories to every city and a lot of them now can make a climate action plan," he said. "Without AI this would not have been possible."

- Oil AI -

But while AI could boost the efficiency and use of renewables, the technology has long been used by oil and gas companies to sift through geological data, identify underground reservoirs and streamline fossil fuel production.

"AI can also have a detrimental impact on climate change from that angle," said Maria Joao Sousa, executive director of Climate Change AI, a nonprofit promoting the use of AI for climate action.

"AI will never be a silver bullet," she said. "Technology itself will never be the solution alone."

W.Huber--VB