The use of AI models such as ChatGPT not only has advantages, but also a significant downside: sustainability. A recent study by The Washington Post and several US universities have shown that large language models such as ChatGPT have a significant impact on our environment, not only because of their high energy consumption, but also because of their high water consumption. In this article, we look at the hidden impact of AI on our natural resources and why it is getting more and more attention.
Energy and water consumption in AI models
To train a model like ChatGPT, immense amounts of energy are used. This is already a known problem, but what is less often discussed is the water use involved. A lot of water is lost both during the training of the models and during daily use. According to the study, the amount of water can even reach half a liter per hundred words generated by ChatGPT. This happens because the servers that provide the computing power to run ChatGPT become extremely hot during the processes and need cooling to keep functioning properly.
Read The Washington Post article here.
Why water?
The data centers that perform the calculations for AI models must be continuously cooled to prevent overheating. Sometimes this is done by air cooling, but water is often used to cool the servers. You might think that the water can easily be collected and reused, but that is often not the case. Due to the heat that is released during cooling, a large part of the water evaporates. This vapour is emitted into the air and is thus lost. In many cases, drinking water is even used for cooling, because it is most effective in reaching the right temperatures. This sometimes leads to problems with the availability of drinking water, especially in regions where these resources are already scarce.
The impact on local water sources
The consequences of water consumption are not limited to data centers. Because drinking water is needed to meet cooling needs, using AI technology can increase pressure on local water sources. Companies near these data centers, as well as local residents, sometimes find that their access to water is being restricted. Especially in areas that are already struggling with water scarcity, this can be a serious problem.
More efficient chips as a solution?
Although companies are working on more efficient chips that use less energy and water, there is no structural solution to the problem yet. While technological progress is making new hardware increasingly sustainable, this does not directly change the water consumption problems of the existing infrastructure. Experts agree that this is a concern that won't go away anytime soon.
Sustainability as a growing concern in AI
By the way, the concerns about the impact of AI go beyond sustainability alone. Themes such as data bias, the rise of deepfakes and disinformation, and the consequences for employment are often mentioned in discussions about AI. The possibility of “unhinged” superintelligence also remains a hot topic. Nevertheless, sustainability is increasingly taking a prominent place in the discussion, especially now that it is becoming clear that AI can put enormous pressure on natural resources such as water.
The Washington Post and various experts are calling for the problem not to be ignored. It is important that we become aware of the dark side of AI and that these challenges are being talked about more and more. If AI developers and policy makers don't take this into account, the future of AI could take a significant toll on the environment and our natural resources.
Conclusion
The popularity of ChatGPT and other major language models has undeniable benefits, but it also has hidden environmental costs. The water consumption of data centers, necessary to remove the enormous heat, contributes to pressure on drinking water sources and can even have negative consequences for local residents and other companies. While technology companies are working on more efficient solutions, sustainability remains a major challenge for the AI industry. With these insights, it is time to take a broader look at the impact of AI and to think about how this technology can be used as sustainably as possible.
.svg)
