Skip to Content

The environmental impact of generative AI models

Pascal Brier
Jul 1, 2024

For today’s World Environment Day, I wanted to focus on generative AI models and their impact on our sustainability goals.

Over the past year, #GenAI has revolutionized the tech landscape, showcasing the incredible potential of human innovation and completely taking the world by storm:

  • The number of new LLMs released in 2023 doubled over 2022
  • These models are getting bigger and more complex, necessitating ever growing computing power (see the graph attached from Stanford University’s AI Index Report that highlights this upward trend).

But inevitably, these larger models also consume more energy and generate more and more CO2 (see the second graph also taken from Stanford’s own research)

  • The first version of Llama 2 with 7B parameters produced an estimated 30 tons of CO2.
  • GPT3 with 175B parameters produced an estimated 502 tonnes.
  • GPT-4 is estimated by some to be at 1.75 trillion parameters (10X GPT-3)… and I’ll let you figure out its environmental impact (no official figures issued).

That being said, there are many signs pointing towards more sustainable and energy-efficient practices in the field of AI! The growing research done in energy efficiency, smaller, optimized models, data quality over quantity. All of these are all steps in the right direction.

With AI models becoming more widely used, it is our responsibility as innovators to diligently monitor the environmental effects of using AI systems and encourage its development into sustainable practices as soon as possible.

Meet the author

Pascal Brier

Group Chief Innovation Officer, Capgemini
Pascal Brier was appointed Group Chief Innovation Officer and member of the Group Executive Committee on January 1st, 2021. Pascal oversees Technology, Innovation and Ventures for the Group in this position. Pascal holds a Masters degree from EDHEC and was voted “EDHEC of the Year” in 2017.