AI might drive down carbon — if its energy use is kept in check

When it comes to the impact of artificial intelligence on the climate, the jury is still out.

The potential for rapidly advancing machine learning programs to help reduce greenhouse gas emissions is dizzying, panelists at POLITICO's AI & Tech Summit said Wednesday. That includes everything from improving forecasts of energy demand and weather to speeding innovation on promising energy and climate technologies like nuclear fusion and direct air capture.

At the same time, questions remain about how much energy-hungry AI programs will add to the climate challenge.

"Using more energy means using more fossil-based fuels," said Jake Loosararian, the CEO of Gecko Robotics, an energy infrastructure monitoring company that uses AI.

But others aren't so certain that the growing use of AI would lead to a surge in planet-warming emissions.

"I've been looking at the literature on this over the past couple of months and my main conclusion is the data are really poor," said David Sandalow, a fellow at Columbia University's Center on Global Energy Policy. He noted that the International Energy Agency found that the data centers that run AI and other internet applications were only responsible for 0.6 percent of global greenhouse gas emissions in 2020.

"In the future, it could be an enormous problem," Sandalow said of emissions from AI. "But actually, it could go down as well."

That uncertain emissions trajectory is due to the potential for improvements to the efficiency of the chips, algorithms and data centers that run AI programs, he argued.

"What we need to do is put in place policy frameworks and incentives to make sure that, as the use of AI grows, AI emissions associated with that technology do not increase," Sandalow said.