Artificial intelligence jeopardizes Microsoft’s climate goals

Artificial intelligence jeopardizes Microsoft’s climate goals
Artificial intelligence jeopardizes Microsoft’s climate goals

Have Microsoft’s greenhouse gas emissions increased by 30% since 2020? Yes, the ChatGPT chatbot answers us, “primarily due to the expansion of their artificial intelligence (AI) and cloud computing service.” And for good reason, Microsoft, like all tech giants, is investing massively in AI. He leads several initiatives like Azure AI, Project Bonsai, Power Platform and works closely with OpenAI, the organization that developed ChatGPT. The latter is therefore well placed to know the ecological impact of AI.

In its sustainable development report published on May 15, Microsoft recognizes that on average, these emissions are up 29.1% compared to 2020. A real boom which essentially comes from Scope3, i.e. , emissions from the entire supply chain, upstream and downstream. These emissions come from “from the construction of new data centers and the associated embodied carbon in building materials, as well as hardware components such as semiconductors, servers and racks,” details the report.

“Before you had to be a geek to use AI”

An announcement which considerably slows down Microsoft’s ambitions. The giant had set itself the objective of carbon neutrality by 2030. “In 2020, we unveiled what we called our carbon moonshot. This was before the explosion of artificial intelligence,” Brad Smith, the group’s president, explains to Bloomberg. “So in many ways the moon is five times further away than it was in 2020, if you just think about our own predictions about the expansion of AI and its power requirements.”

For the moment, Microsoft has not announced that it is reviewing its carbon neutrality objective. He hopes to be able to rely on innovations and says he is “optimistic”. Data centers powered by renewable energies, infrastructures allowing water to not be used for cooling… the announcements are multiplying but so are investments in the development of AI. Above all, usage is exploding. “The barrier to entry has lowered, before you had to be a geek, like me, to use generative AI, now it’s accessible to everyone”deciphers for Novethic Théo Alves da Costa, co-president of Data for good.

The latter estimated that between ChatGPT3 and the new addition, ChatGPT4, CO2 emissions could be multiplied by 7 in the best case. “For 100 million people who use it, we arrive at 2 million tonnes of CO2, the size of a CAC40 box”, he explains. A very big difference in particular due to the fact that ChatGPT4 is multimodal, that is to say that it takes into account text, video, sound… thus adding complexity to ChatGPT3 which was a sort of “search Google +++”.

Resignation of employees

Microsoft is obviously not the only tech giant affected. In this frantic race for AI, Google, Amazon, Meta… are investing massively. If the European Union attempts to regulate this appetite with the publication last March of the AI ​​Act, the limits could come from the energy constraints themselves. In a recent interview, the boss of Meta, Mark Zuckerberg, recognized that the high consumption of AI products could limit its development. To train the algorithms, a data center requires a lot of energy over a short period.

In the meantime, it is the internal revolt that could create a surge. At the beginning of May, several Microsoft employees testified to the media Grist. Two of them decided to resign from their positions, denouncing the fact that the AI ​​developed by Microsoft for the fossil industry accelerates oil extraction. “This work to maximize oil production with our technology undoes all our good work, prolongs the fossil fuel era, and allows for untold emissions,” wrote in an email addressed to Satya Nadella, the general director, Holly Alpine, employed in the group since 2014.

In mid-May, Jan Leike and the co-founder of OpenAI, Ilya Sutskever, announced their resignations, denouncing the fact that “OpenAI’s security culture and processes have taken a back seat”. The two employees headed the Overalignment team, responsible for assessing long-term risks. OpenAI ultimately dissolved this division after their departure.

-

-

PREV Dr Welby retires
NEXT The Mayor accuses the flame party of “smoking” the French