Artificial Intelligence Consumes Energy Voraciously: Major Companies Remain Silent

Generative AI uses 30 times more energy than a traditional search engine, warns researcher Sasha Luccioni, who aims to raise awareness about the environmental impact of this new technology.

For years, this Canadian-Russian researcher, named one of Time magazine’s 100 most influential people in the world in 2024, has been working to measure the emissions produced by programs like ChatGPT and MidJourney.

In an interview with AFP during the All In AI conference in Montreal, Luccioni expressed her concern: "I find the use of generative AI to conduct internet searches disappointing."

Language models behind AI programs require enormous computational power to train on billions of data points, necessitating strong servers. Added to this is the energy consumed to respond to user queries.

Another article: A New Revolution: OpenAI Unveils AI That Can Think!

 
She explained that instead of simply retrieving information, "like a search engine finding the capital of a country, for example," AI programs "generate new information," making the process "more energy-intensive."

According to the International Energy Agency, the AI and cryptocurrency sectors consumed around 460 terawatt hours of electricity in 2022, accounting for 2% of global energy production.

Energy Efficiency in AI

In 2020, Luccioni, a leading researcher in AI’s impact on climate, helped develop CodeCarbon, a tool for developers to calculate the carbon footprint of running a piece of code.

Currently, as the head of climate strategy at Hugging Face, an open-access AI model-sharing platform, Luccioni is working on creating a certification system for algorithms.

Similar to the EPA’s Energy Star program, which rates electronics based on energy consumption, this system will allow users and developers to understand the energy footprint of AI products and make more informed choices.

Tech Companies Hiding the Truth

To develop her tool, Luccioni has been testing it on open-access generative AI models. However, she also seeks to apply it to commercial models from Google and OpenAI, the developer of ChatGPT, though these companies have been hesitant.

Despite Microsoft and Google's commitments to achieving carbon neutrality by the end of the decade, both companies saw significant increases in greenhouse gas emissions due to AI in 2023: a 48% rise for Google since 2019, and a 29% rise for Microsoft since 2020.

Read also: Apple to Launch AI Features Next Month: Here's Which Devices Will Get Them
 

We are accelerating the climate crisis, said Luccioni, calling for more transparency from tech companies.

She added that governments, currently flying blind, need to understand what's inside "datasets and how algorithms are trained."

Producing an Image Consumes as Much Energy as Charging a Phone

Luccioni also highlighted the importance of educating people about what generative AI can and cannot do, and at what cost.

In her latest study, she showed that generating a high-resolution image with AI consumes as much energy as fully charging a mobile phone battery.

As more companies seek to integrate AI into daily life through chatbots, connected devices, and online searches, Luccioni called for "energy sobriety."

"The idea is not to oppose AI," she emphasized, "but to choose the right tools and use them wisely."

Follow this article: The Struggle for Digital Identity: How to Distinguish Between Humans and Machines in the Digital Realm