Posted
Update
Video length: 4 mins.
Article written by
Emmanuel Macron has announced that he wants France to become champion in the field of Artificial Intelligence. Nearly 500 million euros will be invested in the sector to achieve this objective. The opportunity for NOWU, partner of franceinfo, to take stock of the opportunities and impacts of AI in terms of the environment in its column “1 planet, solutions”.
It is difficult to say whether artificial intelligence will make it possible to decarbonize our lifestyles, because it is still too recent a technology. A study by PwC and Microsoft estimates that using AI could reduce global CO2 emissions by up to 4% by 2030. Another study by the Boston Consulting Group (BCG) even suggests a reduction of up to 10% CO2 emissions in 2030. But none has scientific validity, because their methodology remains very vague. But it is already possible to see that artificial intelligence is used in certain areas.
AI to help adapt to climate change
Artificial intelligence can help policies better anticipate present and future natural disasters, which will become more frequent and intense with climate change.
- In the agricultural sector, machines equipped with cameras trained to recognize weeds in a field reduce the use of pesticides by spraying only the weeds.
- In the energy sector, RTE (the entity that manages the electricity network in France) uses AI to better predict and manage electricity flows, in particular to integrate the growing share of electricity produced by renewable energies. The objective of this AI is to facilitate the energy transition.
The carbon footprint of artificial intelligence
Like the digital sector in general, AI pollutes. Firstly because of the manufacture of computer equipment, such as electronic chips, which precisely make it possible to process large quantities of data.
Then because the AI consumes a lot of electricity. When an artificial intelligence is created, it must be trained using a large amount of data, and this training consumes a lot of energy. For example, training the AI for GPT-3 alone consumed 552 tons of CO2, according to the University of California. This is equivalent to 205 Paris-New York round trips by plane.
Finally, these data centers which host the data necessary for AI consume large quantities of water to be cooled. Researchers fear that in the future this consumption will have a worrying impact on water reserves, especially in a context of drought.
Towards solutions to reduce the carbon footprint of AI?
There is an awareness in the scientific world and solutions are emerging to reduce the environmental impact of AI: lighter models are starting to be put in place and more sustainable cooling systems are being tested.
For example, Microsoft immersed a data center in the North Sea and its cooling is ensured by the surrounding cold water.
NOWU it’s the positive medium for getting informed and getting active for the planet! Its mission: to enable young Europeans to become actors in the face of environmental challenges through content that relieves guilt and is focused on solutions.