As the world’s water resources become increasingly scarce, the use of Artificial Intelligence (AI) models like ChatGPT can have a huge and largely unexplored impact on the environment. In the Middle East and North Africa (MENA) region, which is already water-stressed, the water consumption of these models could be devastating if used on a large scale.
A study conducted by researchers from the University of California – Riverside and the University of Texas – Arlington found that each time the ChatGPT AI model answers 20 to 50 questions, it uses up the equivalent of a 500-milliliter bottle of water. Even more advanced models like GPT-3 and GPT-4 would consume even more. Training GPT-3 alone uses as much water as cooling 370 BMW cars or 320 Tesla electric vehicles.
Knowing this, companies like the Dubai Electric and Water Authority (DEWA) are being pressed to consider the potential water use of ChatGPT when introducing AI into its services. Other MENA countries that plan to use the AI model in their services face similar risks.
Edurne Gil de San Vicente, program director at the Dubai-based nonprofit Water Alliance, notes that there is a lack of public disclosure about the amount of water used to power AI models. The MENA region already heavily relies on large-scale AC systems and desalination plants that also use a considerable amount of energy and water.
The hope is that leaders of the region will take the necessary steps to incentivize companies to become more sustainable. The UAE has already committed to reach net-zero emissions by 2050 and is hosting the UN COP28 event in Abu Dhabi this year. The World Resources Institute reports that nine out of the top 10 most water-stressed countries are located in the Middle East.
In addition, the region is beginning to incorporate more sustainable practices and energy efficient Green Data Centres. The Moro Hub , created by the Dubai Electric and Water Authority, is the largest solar-powered data center in the world. Other companies in the region such as Microsoft are investing in more advanced data centers in countries such as Saudi Arabia.
By becoming more aware of these issues and taking the necessary steps, the region can take steps to ensure that the environment will not suffer from the increased usage of AI models. Edurne Gil de San Vicente notes that this investment is critical if the MENA region wants to prevent devastating economic losses due to climate change.
