Montreal — Generative Artificial Intelligence (AI) is revolutionizing industries, but its environmental impact is raising alarms. According to Sasha Luccioni, a Canadian computer scientist and climate expert, generative AI consumes 30 times more energy than a traditional search engine, making it a significant contributor to global emissions.
Luccioni, recognized as one of the 100 most influential people in AI by Time magazine in 2024, has long been working to quantify the carbon footprint of technologies like ChatGPT and Midjourney. Speaking at the ALL IN AI Conference in Montreal, she highlighted the massive energy consumption required to train these large language models and respond to individual queries.
Generative AI vs. Traditional Search Engines:
While traditional search engines like Google or Bing extract existing information from the web, generative AI models create new information. This process requires significantly more computational power, as it relies on analyzing billions of data points in real-time to generate responses. This increased computational demand results in much higher energy consumption.
For example, asking a search engine to identify the capital of a country is a straightforward process, whereas generative AI builds a response from scratch, using vast amounts of energy at both the training stage and individual query level.
AI’s Growing Carbon Footprint:
The rise of AI technologies parallels the growth of the cryptocurrency sector, both of which have been flagged for their environmental toll. The International Energy Agency reports that combined, these sectors consumed 460 terawatt hours of electricity in 2022, representing 2% of global electricity production.
Tech giants like Google and Microsoft have publicly committed to achieving carbon neutrality by 2030. However, both companies saw a significant rise in their greenhouse gas emissions due to AI activities, with Google’s emissions up by 48% from 2019 to 2023 and Microsoft's by 29% since 2020.
Efforts Towards Energy Efficiency:
Luccioni is leading efforts to mitigate AI’s environmental impact. In 2020, she co-created CodeCarbon, a tool that helps developers quantify the carbon footprint of running code. CodeCarbon has been downloaded over a million times, reflecting growing interest among developers in understanding and reducing the energy usage of their code.
Currently, Luccioni is spearheading an initiative at Hugging Face, a platform for sharing open-source AI models, to develop a certification system that rates the energy efficiency of AI algorithms. This system could work similarly to the US Environmental Protection Agency’s energy efficiency ratings for appliances, helping users make informed decisions when choosing AI tools.
However, Luccioni stresses that transparency is crucial. While she has tested her energy measurement tools on open-source AI models, commercial models from companies like Google and OpenAI have been more resistant to participating in such studies.
The Role of Governments:
To address AI’s environmental impact, Luccioni advocates for government intervention. Governments currently lack the information needed to regulate AI technology effectively, including details about how datasets are built and how algorithms are trained. She believes that once transparency is achieved, governments can start enacting legislation to ensure AI technologies are developed and used in an energy-efficient manner.
Energy Sobriety and Public Awareness:
Luccioni also emphasizes the importance of public awareness. In one of her recent studies, she demonstrated that creating a high-definition image with AI consumes as much energy as fully recharging a smartphone. As more companies integrate AI into everyday life — through conversational bots, connected devices, and online searches — understanding AI’s true cost becomes essential.
Her message isn’t to reject AI, but to encourage energy sobriety — using the right tools for the right tasks and optimizing AI applications for better energy efficiency.
Conclusion:
As generative AI continues to grow in popularity, it is critical to consider its environmental impact. Sasha Luccioni warns that without transparency and energy-conscious practices, AI could accelerate the climate crisis. Governments, developers, and tech users alike must make informed decisions to balance AI’s benefits with its ecological footprint.
