A recent American study reveals the energy cost of artificial intelligence. This technological innovation could accelerate climate change if not used wisely.
Artificial intelligence is omnipresent, whether as a personal assistant or as a generator of written and imaged data. An omnipresence that translates into a more than considerable environmental impact. A recent American study highlights these astronomical costs, revealing very high water consumption.
Artificial Intelligence is an Energy Guzzler
A study conducted by the University of California at Riverside relayed by Tom’s Hardware highlights the costs associated with using generative AI. For example, we learn that to generate the equivalent of 100 words, ChatGPT requires the equivalent of three 500 ml bottles of water. A paltry figure, but one that must be put into a global context. According to the Moderator’s BlogChatGPT is reportedly used by nearly 180 million people per month. These figures only correspond to use for generating text, we are not talking about images or videos here.
The same study indicates that the Meta group had to use 22 million liters of water to train its LLaMA-3 model, which is the amount of water needed to grow 2,000 kg of rice, or the rice consumption of 164 Americans per year.
Water is not the only resource being tapped, electricity is also part of the package. Reuters reports that a nuclear power plant has been restarted in the United States to meet Microsoft’s AI energy needs. A direction that contradicts recent statements by Microsoft representative Craig Cincotta. According to him, the company “would be working on methods of cooling data centers that will completely eliminate water consumption”. While data centers do not directly consume water, power plants need it for their operation.
A global problem
The tech industry has quickly seized on artificial intelligence as its new jewel. Every company is coming up with its own innovation, whether it’s a new language model or building a data center. Recently, Elon Musk boasted about the rapid construction of his new AI server, while some associations accuse him of worsening pollution in the region where he has set up shop.
This growing demand is putting pressure on other sectors of the industry such as semiconductors (using rare materials) which are finding themselves on a tight schedule to meet these challenges.
While AI represents a major technological advance, it entails environmental and societal costs that different countries must take into account.
Source: www.frandroid.com
