While artificial intelligence is used in many areas, particularly in environmental issues such as climate change and sustainability, the energy consumption of this technology raises question marks.
According to the US Environmental Protection Agency (EPA), the integration of artificial intelligence into search engines has created the need to perform more calculations and models, and the energy required for each search has therefore increased fivefold.
According to a study by scientists at the University of Massachusetts Amherst, an artificial intelligence model while working on a new language model causes 320 tons of carbon dioxide emissions, and this amount is estimated to be equivalent to the carbon dioxide emitted by five cars during their entire service life. The research highlighted that training GPT-3, one of the artificial intelligence technologies, consumed 1287 megawatt hours of electricity, resulting in 502 tons of carbon dioxide emissions.
Answering questions about the connection between artificial intelligence and the environment, Head of Computer and Informatics Faculty of Istanbul Technical University (ITU), Department Head of Computer Engineering and Vice President of the International Agricultural Informatics Association (Agromatics Society), Prof. DR. Berk Üstündağ said that artificial intelligence can make a significant contribution to environmental technologies and enable the reduction of the carbon footprint and the efficient use of water resources.
Stating that artificial intelligence enables point measurements on topics such as cleaning technologies and air quality, Üstündağ explained that digitalization will gradually develop into a new generation artificial intelligence and that through its use, Turkey can significantly reduce its overall national and international carbon emissions.
Üstündağ said: “The ability of artificial intelligence to study and predict events at the quantum level will accelerate our development of alternative energy systems with a lower carbon footprint, including fusion energy. Therefore, new nuclear energy methods are being introduced without nuclear waste and radioactive pollution.” “Among those who do this, those who use technologies based on artificial intelligence will be in the lead,” has given his assessment.
Regarding the importance of artificial intelligence in agriculture and industry, Üstündağ said:
“Using artificial intelligence, I can create water regulations on a plot-by-plot basis by dividing them into phenological phases (stages of plant development), but a farmer cannot do this by observing them all individually. So I have to serve it.” . If I use artificial intelligence and bring enough digitalization forward, I can increase the rate by 30 percent, which is more.” “I can save a lot. I can produce a total of 25 percent more economically on a basin basis, according to plan. And what is more important: I am not stealing my grandchildren’s water, but rather creating a more ecologically sustainable system and can drive this forward.
“The water footprint of the 20-question study is 500-700 millimeters”
Istanbul Technical University (ITU), Faculty of Civil Engineering, Department of Environmental Engineering, Lecturer and Coordinator of Sustainability Office, Assoc. Dr. Börte Köse Mutlu explained that artificial intelligence infrastructure consumes not only electricity but also water, and the water used directly for cooling or indirectly for power generation is called virtual or embedded water.
Stating that ChatGPT, one of the most widely used artificial intelligence technologies, produces 8.4 tons of carbon dioxide emissions annually, Mutlu shared the following information about the water footprint of artificial intelligence:
“In a 20-question ChatGPT study, water consumption is at least half a liter, about 500-700 milliliters, the same as a medium-sized water bottle. This is the quantity calculated from the user perspective. The actual share of manufacturers in Training models, teaching languages and development processes through ChatGPT is 700,000 per year.” “It is said to use liters of water. Of course, the user has a footprint, but the footprint during its development must be reduced. There.” If the awareness of this is really high, the manufacturers have started to develop strategies.”
Stressing that even a simple query on ChatGPT creates a 100 times larger footprint than a search engine search, Mutlu said that artificial intelligence is not yet at the forefront of carbon emissions generated by sectors, but by 2030, the infrastructures that are created by Artificial intelligence applications will be used in Europe to reduce energy consumption by 30 percent. He pointed out that an increase is to be expected.
Mutlu pointed out that the cooling systems used to reduce the heat effect when operating models and training artificial intelligence languages cause significant energy consumption, mentioning that 40 percent of the energy consumption of artificial intelligence is spent on cooling systems .
Stressing that alternative energy sources should be used so that the energy used in the production and use of artificial intelligence does not cause pollution, Mutlu said: “Indeed, if we meet the consumption of these information technologies with green, clean and renewable technologies.” Energy “Both the carbon footprint and the direct water footprint decrease,” he said.
“We can use artificial intelligence in biodiversity”
Mutlu shared the view that waste is one of the problems threatening the environment and performance can be increased by using artificial intelligence in the classification and recycling of waste, explaining that by using the visual perception system in the Segregation and grouping of waste future decisions can be made. When historical data is used, waste minimization and recycling of the elements it contains can be facilitated.
Mutlu predicted that incorporating artificial intelligence into climate scenarios and modeling studies could make climate models much closer to reality, summarizing the other benefits this technology can bring to the environment as follows:
“As we use artificial intelligence in smart city and transportation applications, such as determining routes and controlling traffic and lanes, we will reduce direct CO2 emissions through the proliferation of autonomous vehicles. After that, it will help us collect data and make sustainable decisions. Agriculture, smart irrigation, fertilization, use of pesticides, herbicides and biocides. Biodiversity is also very important and a topic that we do not directly relate to artificial intelligence bring, but the data of both living species and forest and landscape products need to be collected, monitored and sometimes certain decisions need to be made. Saying this, we can use artificial intelligence in our actions regarding biodiversity. “We can stop it from decreasing and sometimes disappearing, and we can take immediate action against deforestation.”