![\"\"](\"https:\/\/www.enerjigazetesi.ist\/wp-content\/uploads\/2021\/02\/yapay-zeka-ve-enerji-verimliligi.jpg\")
intelligence, a technology that relies on processing huge amounts of data to recognize patterns. When we use the recommendation algorithm of our favorite streaming service, we usually don’t realize the gigantic energy consumption behind it. The billions of operations needed to process the data are typically carried out in data centers. All these computations consume a tremendous amount of electric power. Although data centers heavily invest in renewable energy, a significant part of the power still relies on fossil fuels. The popularity of AI applications<\/strong> clearly has a downside: the ecological cost.<\/strong><\/p>\nTo get a better understanding of the total footprint<\/strong>, we should take two factors into account: training and inference. First, an AI model needs<\/strong> to be trained by a labeled dataset<\/strong>. The ever-growing trend<\/strong> toward the use of bigger datasets for this training phase causes an explosive growth in energy consumption. Researchers from the University of Massachusetts<\/strong> calculated that during the training of a model for natural language processing,\u00a0284 metric tons<\/strong>\u00a0of carbon dioxide is emitted. This is equivalent to the emission of five cars during their entire life span, including construction. Some AI models developed by tech giants \u2014 which are not reported in scientific literature \u2014 might emit at a greater magnitude.<\/p>\n The training phase is just the beginning of the AI model\u2019s life<\/strong> cycle. Once the model is trained, it is ready for the real world: finding meaningful patterns<\/strong> in new data<\/strong>. This process, called inference, consumes even more energy. Unlike training, inference is not a one-off. Inference takes place continuously. For example, every time a voice assistant is asked a question and generates an answer, extra carbon dioxide is released. After about a million inference events, the impact will surpass that of the training phase. This process is unsustainable.<\/p>\n