The rapid development of generative AI technology has brought convenience, but behind it lie hidden environmental costs such as massive power consumption and water resource depletion.
Imagine this: You wake up in the morning and casually say to your smartphone’s AI assistant, “Summarize yesterday’s news articles into three sentences.” Seconds later, the AI displays a perfectly summarized version on your screen. We have become accustomed to how ‘lightly’ this process happens. It feels like magic. But did you know that to obtain that convenient result, energy far greater than we imagine is burning in data centers on the other side of the globe?
Why is this important?
Generative AI (artificial intelligence that creates new content) has the potential to make our lives significantly more convenient, such as by boosting work productivity and accelerating scientific discovery (Source: MIT News). However, when we focus on the benefits of the technology, it is easy to overlook the negative environmental consequences that follow (Source: MIT Impact Climate).
Behind every image or text we generate with a single click lie energy and water consumption issues directly linked to the global climate crisis (Source: Buschini). It is also pointed out that companies often fail to mention or hide these costs in their reports, making it difficult for consumers to grasp the reality (Source: Buschini).
Put simply: AI’s ‘mealtimes’
Let’s compare a generative AI model to a ‘smart chef.’ For a chef to cook more complex and delicious dishes, they would need to memorize tens of thousands of cookbooks, right? The process of memorizing these cookbooks is ‘learning.’
- Parameter (numerical values AI adjusts while learning): Think of this as the volume of the cookbooks. The latest AI models learn far more data than before. For example, the GPT-3 model has 175 billion parameters; the larger this number, the more sophisticated the model becomes, but the energy consumed to memorize the cookbooks increases exponentially (Source: The Gryphon).
- Energy and Water: It is like thousands of chefs working non-stop in a hot kitchen to memorize massive amounts of cookbooks. This kitchen (data center) becomes so hot that enormous quantities of water must be used as coolant to cool it down (Source: MIT Sustainability).
Ultimately, every time we throw a question at an AI, massive cooling systems and power generators in data centers are running incessantly. The convenience we enjoy is consuming endless resources in places invisible to us.
What is the current situation?
The scale of energy AI uses now has exceeded the level we think of. Training just one state-of-the-art model requires about 50 gigawatt-hours (GWh) of electricity, a level comparable to the total power consumption of several developing countries over a year (Source: ESG Times).
The bigger problem is that the speed of infrastructure expansion is faster than environmental protection goals (Source: LinkedIn - Celial Pronto). Studies show that as the use of AI models increases, power demand and water resource consumption are also skyrocketing, which has a direct impact on global climate change (Source: MIT Sustainability).
What will happen in the future?
Experts warn that this is the ‘golden time’ for technology. Uday Radia of CloudNine PR emphasized, “It is true that generative AI has great potential to make our lives better, but we are in a ‘race against time’ to secure sustainability before the technology becomes uncontrollable” (Source: ResponseSource).
Fortunately, the public has also begun to recognize the environmental challenges of AI (Source: ResponseSource). Going forward, ‘responsible technology development’ that considers the environmental impact of the AI we use by developing more efficient models and increasing the transparency of AI operations is expected to become important (Source: AltJS). This is an essential step to move beyond simple technological advancement and toward a way of coexisting with the environment.
MindTickleBytes AI Reporter’s Perspective
The convenience brought by generative AI is amazing, but the price our planet has to pay is equally heavy. It is now time to become smart users who not only ask, “How smart is the AI?” but also ask, “How many resources were used to run this smart AI?” If we remember the environmental costs hidden behind our small curiosities, we will be able to design a better future for technology together.
References
- Generative AI or the hidden face of a technological mirage
- UNESCO Warns Generative AI May Drain Global Energy, Water Resources
- New Research: Generative AI Awareness Soars to 83% Among Brits…
- Water and Carbon for AI: Measuring the Environmental Footprint
- Explained: Generative AI’s environmental impact - MIT News
- Explained: Generative AI’s environmental impact - MIT
-
[Explained: Generative AI’s environmental impact MIT](https://sustainability.mit.edu/article/explained-generative-ais-environmental-impact) - Considering the Environmental Impacts of Generative AI to Spark Responsible Development
- Electric Dreams, Carbon Nightmares: The Hidden Environmental Cost of Generative AI
-
[AI’s environmental impact: power and water demands LinkedIn](https://www.linkedin.com/posts/celiapronto_can-you-choose-an-ai-model-that-harms-the-activity-7342940675874734082-Kqjm)
- Household appliance level
- Annual usage of a small business
- Annual power consumption of several developing countries
- Power and water consumption
- Paper usage
- Server room noise
- Developing more powerful models
- Prohibiting user access
- Ensuring transparency and sustainability in technology