The promise of green artificial intelligence collides with the binge of uses

In any case, the future of the climate does not seem more stable since the advent of ChatGPT, Gemini, Claude, and other chatbots.”

Benjamin Ninassi, a researcher at Inria, is particularly vigilant about the claims of industrialists. “Before we can say that a technology will have a positive impact, we need to apply a rigorous calculation method,” he insists. The Ademe (Agency for the Environment and Energy Management) has developed a methodology for evaluating scenarios to compare the impact of a technological solution by integrating all equipment, beyond just the usage phase.

A new use of AI inevitably leads to increased consumption

Because this is where one of the blind spots of the debate resides.

“It’s very easy to highlight a solution that will reduce greenhouse gas emissions, but which can increase other types of pollution,” the researcher points out. “There will potentially be a transfer of impact.”

The analysis must take into account all components (water, raw materials, energy) and not be limited to the usage phase.

The main overflow remains the rebound effect. Any improvement in efficiency reduces the marginal cost of use, which stimulates demand, generates new behaviors, and ultimately cancels out, or even surpasses, the initial savings.

“This is exactly what we have observed with digital technology for forty years,” recalls Benjamin Ninassi. “With each technological generation, the promise was to consume less.”

To read more, click here