Preko Zemljinih Granica

A Slovenian columnist has raised concerns over the water consumption linked to artificial intelligence systems, highlighting how even small interactions with AI models can have a substantial environmental impact. The issue centers around the amount of water required to power and cool the servers that run these technologies, particularly large language models such as ChatGPT and GPT-3. According to the writer, one query and response from ChatGPT, which consists of 100 words, consumes approximately 517 milliliters of water. This figure includes both the water used for cooling the servers and the water necessary to generate the electricity that powers them. The columnist points out that while the direct water usage per interaction appears minimal, just 0.3 milliliters, the scale of AI usage dramatically increases this number. With ChatGPT reportedly receiving nearly a billion queries daily, the annual water consumption could range from 4.2 to 6.6 cubic kilometers. For context, the average person in Slovenia uses between 40 and 60 cubic meters of water annually. If scaled up, the water demand for AI operations alone would surpass that of millions of individuals. Furthermore, the writer emphasizes that these figures do not account for additional water loss through evaporation or leakage within data centers. The production of electricity needed to sustain AI infrastructure also contributes significantly to the overall water footprint. For instance, the creation of the GPT-3 model required approximately 700,000 liters of water for cooling and 5.4 million liters for generating the electricity that powered the servers. As new AI models continue to emerge and older ones remain operational, the cumulative effect on water resources becomes increasingly pronounced. The article references a United Nations University study suggesting that the global water consumption associated with AI could reach unprecedented levels by the end of the decade. While the exact projections remain under discussion, the trend indicates a growing concern over the sustainability of AI technologies. Environmentalists argue that the energy-intensive nature of AI systems, combined with their reliance on water for cooling and power generation, poses a serious challenge to global water conservation efforts. In addition to the direct water usage, the environmental impact extends beyond just the immediate consumption. The process of producing and distributing the electricity that fuels AI systems involves extracting and processing raw materials, which often requires significant water inputs. These factors contribute to a broader ecological footprint that encompasses not only water but also land use, pollution, and greenhouse gas emissions. As awareness of these issues grows, there is increasing pressure on tech companies and researchers to develop more sustainable practices. Some initiatives focus on improving the efficiency of data centers, using renewable energy sources, and implementing advanced cooling techniques that reduce water dependency. However, the transition to greener AI solutions will require coordinated efforts across industries and governments to ensure that technological progress does not come at the expense of environmental health.