Source URL: https://news.slashdot.org/story/25/04/24/1556239/even-the-us-government-says-ai-requires-massive-amounts-of-water?utm_source=rss1.0mainlinkanon&utm_medium=feed
Source: Slashdot
Title: Even the US Government Says AI Requires Massive Amounts of Water
Feedly Summary:
AI Summary and Description: Yes
Summary: The Government Accountability Office (GAO) report highlights significant environmental impacts of generative AI systems, particularly in terms of water usage and energy consumption for cooling data centers. Companies’ lack of transparency regarding resource usage raises concerns about sustainability and compliance in the AI domain.
Detailed Description: The recently released report from the Government Accountability Office (GAO) sheds light on the environmental costs associated with generative AI systems, particularly focusing on their substantial water and energy usage. Here are the major points from the report:
– **Water Consumption**: The report reveals that generative AI systems are responsible for an astonishing water consumption of over 1.1 million gallons to support 250 million daily queries, which presents a significant environmental challenge.
– **Energy Demand**: Cooling data centers, where these AI models are housed, consumes between 100 and 1000 megawatts of power, accounting for approximately 40% of the total energy consumption in these facilities. This figure is projected to increase with rising global temperatures, emphasizing the urgent need for sustainable practices in infrastructure.
– **Geographical Variability**: Water usage for these systems is highly influenced by geographical factors, suggesting that different areas will have varying impacts on local resources and carbon emissions.
– **Carbon Emissions**: The report offers concrete examples of the carbon footprint for notable generative AI models, showing that Meta’s Llama 3.1 405B model generated 8,930 metric tons of carbon, while Google’s Gemma2 and OpenAI’s GPT-3 generated significantly less at 1,247.61 metric tons and 552 metric tons, respectively.
– **Cost of Searches**: It was found that generative AI searches are approximately ten times more expensive than standard keyword searches, raising questions about the cost-effectiveness of these technologies relative to their environmental impact.
– **Transparency Issues**: The GAO report calls attention to the lack of transparency regarding resource usage among companies developing these AI systems, criticizing the systems as “black boxes.” This highlights ongoing accountability issues and regulatory implications that may arise as the industry evolves.
The findings from this GAO report are crucial for professionals in security, compliance, and infrastructure as they underscore the importance of integrating sustainability considerations into the AI development lifecycle, ensuring regulatory compliance, and promoting transparency within the industry. The environmental footprint of generative AI raises significant issues regarding governance and responsible resource management that are imperative for future advancements.