Firstly, what is Artificial Intelligence (AI)? According to UNESCO, AI is a group of technologies that enables machines to mimic and imitate human intelligence. This can span from problem solving, perception, language interaction, and even creativity. AI is an extremely powerful tool and has become fundamental for many businesses and organizations today.
Many AI models, such as ChatGPT, have data centres. Since AI doesn’t have to take a break unlike the human brain, these data centres are constantly running on energy. These data centres rely on electricity to develop and run AI models. ChatGPT is considered a large-scale AI model, as you can maybe tell from its popularity nowadays. I personally have a lot of peers who use ChatGPT to assist them with their schoolwork.
Now, how does the scale of AI affect the environment? Due to the staggering amount of electricity these data centres need to run, it can directly affect greenhouse emissions, which further aggravates climate change. According to Scientific Computing World, the estimated carbon cost of training an AI model is 626,155 lbs. That’s the equivalent of 125 round-trip flights from New York to Beijing. The more AI models develop and become complex, the energy needed to run these AI models increases.
These data centres also need water to cool down the hardware used to develop and run these AI models. To put that in perspective, one ChatGPT prompt uses 519 millilitres of water, which is the equivalent of one water bottle (University of California, n.d., as cited in AIAAIC, 2023). Due to the large-scale nature of these data centres, it requires plenty of water to cool down. Cooling data centres are usually expensive and inefficient, especially in hotter climates. This becomes a bigger issue when a quarter of the world lacks access to clean and safe drinking water (United Nations, n.d.). As the demand for AI grows, it amplifies the strain on electricity and water resources.
While the impact of AI on the environment is significant and undeniable, efforts have been made to make it more sustainable. For instance, Google has made an effort to strictly use renewable energy sources, such as solar and wind, to power its AI data centres (Sanderson, 2024). A lot of these data centres consume a significant amount of power. It's important that organizations look towards renewable energy sources, particularly to reduce their environmental impact and to promote long-term sustainability.
Although the world has a long way to go in minimizing the environmental impact of AI, there are steps you can take to help reduce its effects. Staying conscious of AI tools is a great place to start. Limit any unnecessary usage of AI and be mindful of how you use those tools. Stay informed about AI’s environmental impact and advocate where you can. Individual actions may seem small, but collectively we have the power to drive meaningful change and a sustainable world.
References:
AIAAIC - ChatGPT consumes 500 ml of water per 5-50 prompts. (2023, September). https://www.aiaaic.org/aiaaic-repository/ai-algorithmic-and-automation-incidents/chatgpt-consumes-500-ml-of-water-per-5-50-prompts
Sanderson, C. (2024, December 11). Google to spark $20bn green power spree in AI data centre “rethink” pact. Recharge. https://www.rechargenews.com/energy-transition/google-to-spark-20bn-green-power-spree-in-ai-data-centre-rethink-pact/2-1-1751988
The true cost of AI innovation | Scientific Computing World. (n.d.).
https://www.scientific-computing.com/analysis-opinion/true-cost-ai-innovation
UNESCO. (n.d.). Artificial Intelligence. https://www.unesco.org/en/artificial-intelligence
United Nations. (n.d.). Water | United Nations. https://www.un.org/en/global-issues/water
