Apr 18, 2023 | Posted by MadalineDunn
Data centers are already known for guzzling enormous amounts of water, so as AI enters its "gold rush" era, how will it affect water consumption? Well, a new study by researchers from the Universities of Colorado Riverside and Texas Arlington found that training large AI models like OpenAI's ChatGPT and Bard, uses an incredible amount of water.
The researchers discovered that training GPT-3 at Microsoft's data centers would result in the facilities consuming 700,000 liters, and that's a conservative estimate based on the assumption that the data centers are high efficiency. Comparatively, less efficient data centers could suck up 4.9 million liters. And, it's not just any old water that's required to quench the thirst of these data centers, it's fresh water, to prevent corrosion and bacterial development, and due to the humidity of the rooms where they're housed.
With the Chatbot becoming increasingly popular, especially against the backdrop of widespread drought, concerns are being raised about water supplies. Further, with AI's rapid evolution, the researchers predict that as new models emerge, water requirements will rise.
Speaking about this, the researchers said that AI models' water footprint can "no longer stay under the radar." "Water footprint must be addressed as a priority as part of the collective efforts to combat global water challenges," they added.
In terms of recommendations of how to navigate this issue, the researchers recommended that AI model developers and data center operators "be more transparent," which would help consumers make more informed decisions about how they use the technology: "When and where are the AI models trained? What about the AI models trained and/or deployed in third-party colocation data centres or public clouds?"
It's important to note that currently, one in ten people on the planet lack access to clean water, and predictions are that by 2025, two-thirds of the world's population will be faced with water shortages without drastic intervention.
As more is known about the rate and quantity of AI water consumption, and the increasing fears surrounding its development, some believe it could be time to take our foot off the gas.
