Experts claim that the advancement of artificial intelligence has caused a spike in energy consumption due to the fact that most people used to rely on the Google search engine, which operates on less powerful CPU processors compared to the GPUs used in AI centers.
Recent research indicates that the training and progress of artificial intelligence have consumed an amount of electricity equivalent to the consumption of a typical family for 23 years.
On November 30, 2022, when OpenAI launched ChatGPT, the International Energy Agency (IEA) reported that this event marked the beginning of a significant increase in the consumption of energy resources. GPU centers consumed between 1% and 1.3% of global electricity in 2022.
Additionally, according to Google data, the owner of the Gemini AI model, the multinational company increased its energy consumption by 16.2% in 2023.
This means that to get a response from ChatGPT, graphics processing units need three times more energy consumption than a normal search on Google, according to data from the Institute of Engineering of Spain (IIE).
For example, the energy consumed just for training a model like GPT-3 is comparable to the consumption of a Spanish household for 23 years (78,437 kWh).
On the other hand, new chatbots require water to operate, both for electricity production and for equipment cooling. In data centers, the environment is sprayed to prevent equipment overheating.
The United Nations Environment Programme (UNEP) warns that globally, AI-related infrastructure could consume six times more water than a country like Denmark by 2027.
Not only should energy consumption be considered, but also the waste they produce. These data centers generate components that often contain hazardous substances such as mercury and lead. Manufacturing a 2-kilogram machine requires 800 kilograms of raw materials.
A study by the University of Cambridge and the Urban Environment Institute of the Chinese Academy of Sciences, published in Nature Computational Science, states that the waste generated from technology will increase from 2,600 tons in 2023 to 2.5 million tons in 2030, equivalent to discarding between 2.1 and 13.3 billion units of iPhone 15 Pro.
Have you visited our YouTube channel yet? Subscribe now!
