“In the event you’re doing a particular application, like looking through email … do you really want these big models which might be able to anything? I might say no,” Luccioni says. 

The energy consumption related to using AI tools has been a missing piece in understanding their true carbon footprint, says Jesse Dodge, a research scientist on the Allen Institute for AI, who was not a part of the study. 

Comparing the carbon emissions from newer, larger generative models and older AI models  can also be essential, Dodge adds. “It highlights this concept that the brand new wave of AI systems are rather more carbon intensive than what we had even two or five years ago,” he says. 

Google once estimated that a median online search used 0.3 watt-hours of electricity, equivalent to driving 0.0003 miles in a automobile. Today, that number is probably going much higher, because Google has integrated generative AI models into its search, says Vijay Gadepally, a research scientist on the MIT Lincoln lab, who didn’t take part in the research. 

Not only did the researchers find emissions for every task to be much higher than they expected, but they found that the day-to-day emissions related to using AI far exceeded the emissions from training large models. Luccioni tested different versions of Hugging Face’s multilingual AI model BLOOM to see what number of uses can be needed to overtake training costs. It took over 590 million uses to succeed in the carbon cost of coaching its biggest model. For highly regarded models, equivalent to ChatGPT, it could take just a few weeks for such a model’s usage emissions to exceed its training emissions, Luccioni says. 

It’s because large AI models get trained only once, but then they will be used billions of times. In line with some estimates, popular models equivalent to ChatGPT have as much as 10 million users a day, lots of whom prompt the model greater than once. 

Studies like these make the energy consumption and emissions related to AI more tangible and help raise awareness that there’s a carbon footprint related to using AI, says Gadepally, adding, “I might find it irresistible if this became something that customers began to ask about.”

Dodge says he hopes studies like this can help us to carry corporations more accountable about their energy usage and emissions. 

“The responsibility here lies with an organization that’s creating the models and is earning a profit off of them,” he says.