人们常关注人工智能（AI）对能源的消耗，但较少注意到其对水资源的影响。实际上，AI在回答每一个问题时，背后依赖的是大型数据中心运行的复杂计算系统，而这些数据中心在运行过程中会产生大量热量，需要冷却系统维持设备正常工作。目前主流的冷却方式之一是使用水冷技术，即通过蒸发或循环冷却水来散热。根据2023年一项由美国大学研究人员发表的研究估算，一次典型的AI问答（如使用ChatGPT类模型）平均消耗约500毫升至1升的水。这个数字看似微小，但考虑到全球每天数以亿计的AI请求，累积的用水量相当可观。例如，训练一个大型语言模型（如GPT-3）可能消耗数百万升水，而日常推理（即用户提问后AI生成答案）也会持续产生用水需求。值得注意的是，不同地区数据中心的冷却效率和水源条件差异很大，因此实际耗水量会因地理位置、气候和冷却技术而异。随着AI应用的普及，科技公司正积极研发更节水的冷却方案，如采用空气冷却、液冷闭环系统或选址在气候凉爽的地区，以减少对水资源的压力。总之，AI并非“无痕”技术，其环境足迹包括电力与水资源消耗。提升公众对此的认知，有助于推动更可持续的AI发展路径。
While much attention is paid to the energy consumption of artificial intelligence (AI), its water usage is often overlooked. In reality, every AI response relies on complex computing systems housed in large data centers, which generate significant heat during operation and require cooling systems to function properly. One common cooling method involves water—either through evaporative cooling or recirculating water systems.According to a 2023 study by researchers at U.S. universities, a typical AI query (such as one made to a ChatGPT-like model) consumes approximately 500 milliliters to 1 liter of water. While this amount may seem small per query, it adds up significantly given the billions of AI requests processed globally each day. For instance, training a large language model like GPT-3 can consume millions of liters of water, and ongoing inference—when users ask questions and receive AI-generated answers—continues to drive water demand.It’s important to note that actual water consumption varies widely depending on the location, climate, and cooling technologies used by data centers. As AI adoption grows, tech companies are actively developing more water-efficient solutions, such as air cooling, closed-loop liquid cooling, or siting data centers in cooler climates to reduce water strain.In summary, AI is not a “zero-footprint” technology—its environmental impact includes both electricity and water use. Raising awareness of this issue can help steer AI development toward more sustainable practices.
原创文章，作者：admin，如若转载，请注明出处：https://ktez.cn/12279.html