The huge impact of AI creators to the environment

The explosion of artificial intelligence (AI) is causing increasing environmental concerns. This is also one of the top issues discussed at the Global Conference AI Action Summit is scheduled to take place in Paris on February 10. Ahead of the conference, AFP French news agency made some important statistics about AI.

Each request sent to chatgpt, chatbot has the ability to create a variety of feedbacks for questions in natural language, consuming 2.9 whites. This figure is 10 times the amount of electricity consumed for one search on Google, according to the International Energy Agency (IEA). Openai, American company developing chatgting, stated that this chatbot currently has 300 million users weekly, making a total of one billion requirements per day.

In addition to chatgpt, chatbot introduces who created to the public in 2022, there are thousands of other chatbot. The poll of the IFOP survey company (France) shows that 70% of people 18-24 years in this country use who created birth. In the United States, the poll of Morning Consult shows that 65% of teenagers are 13-17 years old to use who generate, this figure in the common population is nearly 50%.

Whoever creates birth can not operate without data centers and enormous calculations. In 2023, data centers accounted for nearly 1.4% of the global electricity consumed, according to Deloitte’s research. However, with many large investments pouring into the creator, this rate is expected to reach 3% by 2030 – equivalent to 1,000 TWH. Deloitte said that this figure is equivalent to the annual consumption of France and Germany.

IEA predicts the electricity consumption of data centers in 2026 will increase by 75% compared to 2022, reaching 800 TWH. US Consulting Company Gartner said that the huge demand for electricity is synonymous, 40% of the data center for AI applications can face electricity shortage by 2027.

The training of one of the large language models (LLM) provides energy for chatbot produces about 300 tons of CO2 greenhouse gases, according to the research group at the University of Massachusetts Amherst in 2019. With 125 round -trip flights between New York and Beijing.

Two years later, the research team at Oxford University gave 224 tons for an Openai GPT-3 model training session. Developers must train thousands of models to promote progressive technology.

Despite these estimates, it is very difficult to assess the total number of greenhouse gas emissions. Many experts and organizations have pointed out the lack of information on how the models are produced and lack of global measurement standards.

Not only energy, whoever creates also consumes water, especially to cool the computer hardware. GPT -3 needs about 0.5 liters of water to create 10-50 response, according to the prudent estimate of the California Riverside University and Texas Arlington University.

Overall, AI’s water demand is expected to reach 4.2 – 6.6 billion m3, according to a 2023 study. This figure is 4-6 times the amount of water consumed every year.

About 2,600 tons of electronic waste such as graphics cards, servers and memory chips arising from AI applications born in 2023, according to research in the magazine Nature Computational Science. The team predicts this figure will reach 2.5 million tons by 2030 if the current trend continues and has no measures to limit waste. That figure will be equivalent to about 13.3 billion smartphones discarded.

Like many computer hardware, AI device contains chips that also require rare metal to produce. The exploitation of this metal, usually in Africa, may include serious pollution processes.