The popularity and usage of AI systems like ChatGPT is growing and people are using this AI Chatbot for various tasks, from writing plays to creating websites.
These advanced technologies are not without their environmental costs, particularly in terms of water and electricity consumption.
AI systems, including ChatGPT, require significant amounts of water and energy to operate. This is primarily because the data centers where these AI systems are trained need a lot of water to cool their servers. It’s an essential part of digital water consumption that often goes unnoticed.
For instance, a single conversation with ChatGPT can consume about 50 cl of water, roughly equivalent to a small plastic bottle. With ChatGPT having nearly 1.5 billion users monthly, the cumulative water usage is substantial.
Moreover, data centers, which are crucial for AI training, account for nearly 1% of global energy consumption.
In the future, this figure is expected to increase. The carbon dioxide (CO2) emissions from these centers vary depending on the source of electricity in their base country, whether it’s coal, gas, or renewable energy.
Interestingly, research funded by Microsoft and the Allen Institute for Artificial Intelligence has shown that by strategically choosing the location for AI training, it’s possible to reduce CO2 emissions by up to 75%.
This approach, known as “following the sun,” involves changing AI training locations throughout the day to continuously utilize solar energy, thereby optimizing energy use and minimizing environmental impact.
For example, Bloom, an open-source AI similar to ChatGPT, produced about 25 tonnes of CO2 during its training phase, even though most of the energy used was from nuclear power, which is carbon-free.
In comparison, GPT-3, the model behind ChatGPT, has an estimated carbon footprint 20 times higher, equivalent to around 300 round trips from Paris to New York by plane.
Despite these environmental concerns, AIs like ChatGPT can play a significant role in combating global warming. They can assist in predicting extreme weather events or optimizing industrial processes to reduce CO2 emissions.
The challenge for AI developers now is transparency. Most publicly available AI models do not disclose their training locations or the carbon cost of using them.
When users have access to this information, they can make more informed choices. For example, for basic queries, opting for a simple browser search, which consumes less power than querying ChatGPT, can be a more environmentally friendly choice.
While AI systems like ChatGPT offer numerous benefits, it’s crucial to be aware of their environmental impact and make informed choices to minimize their carbon footprint.
