Nuclear Power Fails to Quench AI's Thirst for Water, Says Constellatio
training
| Source: Mastodon | Original article
Nuclear power won't fix AI's water usage problem. Constellation Energy needs 73M gallons daily.
Constellation Energy's recent request to withdraw 73 million gallons of water daily from the Susquehanna River for its nuclear operations has sparked concerns about the water consumption issue in AI. This development is particularly noteworthy given the significant water footprint of AI models, as highlighted in recent research. Training chatbots like ChatGPT requires substantial amounts of water, equivalent to filling large pools.
The issue of water consumption in AI is multifaceted and cannot be solved by simply switching to nuclear energy. While nuclear power itself does not directly consume large amounts of water, the cooling systems used in nuclear plants do. Furthermore, the IT infrastructure supporting AI operations is a significant contributor to water usage. As the demand for AI continues to grow, it is essential to address the water footprint of these technologies.
As we move forward, it is crucial to monitor the development of more water-efficient AI systems and infrastructure. Researchers and industry leaders must prioritize innovative solutions to reduce the water consumption associated with AI. With the growing awareness of AI's environmental impact, we can expect to see increased efforts to mitigate its effects on the environment, including its water footprint.
Sources
Back to AIPULSEN