Study Reveals Training ChatGPT Consumes As Much Water As A Nuclear Reactor

A new study titled “Making AI Less Thirsty” reveals that a single conversation you have with ChatGPT amounts to spilling a half-litre bottle of water on the ground.

As for the energy consumption of ChatGPT training itself? The researchers say you could fill a nuclear reactor’s cooling tower with water – that’s how much fresh water was consumed to cool the server farms that ran ChatGPT in its training. In terms of energy, you could build batteries for 320 Tesla cars for the same amount it takes to train a software like this.

According to researchers from the University of Colorado Riverside and the University of Texas Arlington, training ChatGPT consumed  185,000 gallons (700,000 liters) of water – and this is one large language model (LLM) in a multitude of them, including Google’s Bard AI.

The fact that training LLMs and other generative AI consume energy is not news but many did not know how much freshwater is involved in the process as well. 

Gizmodo put together a straightforward explanation of how chatbots use water. From their report:

“When calculating AI’s water consumption, the researchers draw a distinction between water “withdrawal” and “consumption.” The first example is the practice of physically removing water from a river, lake, or other source, while consumption refers specifically to the loss of water by evaporation when it’s used in data centers. The research on AI’s water usage focuses primarily on the consumption part of that equation, where the water can’t be recycled.

Anyone who’s spent a few seconds in a company server room knows you need to pack a sweater first. Server rooms are kept cool, typically between 50 and 80 degrees Fahrenheit to prevent equipment from malfunctioning. Maintaining that ideal temperature is a constant challenge because the servers themselves convert their electrical energy into heat. Cooling towers like the ones shown below are often deployed to try and counteract that heat and keep the rooms in their ideal temperature by evaporating cold water.”

You can read the research on the water consumption of ChatGPT and other LLMs here.

The researchers conclude the study by saying there needs to be an increased transparency of AI models’ water footprint, not just carbon footprint, in order to create sustainable AI. If you ask ChatGPT if it consumes water, it will say that, since it doesn’t have a physical presence, it can’t do so. Evidently, that’s not the case here.

Also read: OpenAI Launches AI Text Declassifier, A Tool To Check If Text Was Written By Humans