Is AI bad for the environment?
Most people don’t think twice when they ask ChatGPT a question or to summarize an email but they should. AI is working hard for us, but our planet is working even harder. Today we take an in-depth look at why AI, specifically generative artificial intelligence, is bad for the planet.
Quick summary: How is AI bad for the environment?
-
Generative AI uses huge amounts of power: from the data centers, training the model, to the end user interacting with the AI tool.
-
Water is needed to build the data centers, and it’s used in operations to cool down the AI servers. The high water consumption can affect local environments, and ecosystems.
-
The data centers produce a lot of electronic waste, which often contains toxic substances like lead and mercury.
-
The microchips and hardware used to power AI are made using rare elements and minerals, which are often mined unsustainably.
Why is AI bad for the environment?
Generative AI uses huge amounts of energy
The data centers housing generative AI require large amounts of electricity: from when they are built, to training the models, and for operating.
While data centers have been around for a long time, because of the boom of gen AI, they are rapidly increasing in number and being built. An issue with the rise in these centers is that generative AI is very power-demanding.
“A generative AI training cluster might consume seven or eight times more energy than a typical computing workload,”
says Noman Bashir, a Computing and Climate Impact Fellow at MIT Climate and Sustainability Consortium, a postdoc in the Computer Science and Artificial Intelligence Laboratory, and the lead author of The Impact Paper. According to Bashir, the demand for data centers is so high that they cannot be run in a sustainable way. So for now, most of the power that they run on comes from burning fossil fuels, which leads to increased carbon dioxide emissions.
The need for excessive electricity doesn’t stop there. AI models then have to be trained and constantly improved. To do this, companies increase the size of the training datasets and increase the computation used to process the data. Once the model is trained and we (the user) start to ask it questions, to summarize emails, or summarize information, the computing hardware that performs this also consumes significant amounts of energy.
There’s still limited data on the carbon footprint of one generative AI query, for example, when you ask ChatGPT a question, but it’s estimated to be four to five times higher than a query in a search engine like Google.
With this in mind, and considering how normal it is to now search in AI tools like ChatGPT or Gemini you can only imagine how much more power is being used globally.
High water consumption
In addition to being power-heavy, AI uses a lot of water. The data centers need water during construction, and when they are operational, water is required to cool down the hardware used for training and deploying the generative AI models. Because these data centers use a lot of water, this can put pressure on municipal water supplies and disrupt ecosystems.
The hardware
The raw materials needed to build the hardware, like the batteries and semiconductors, use rare earth elements and metals. Very often the mining of rare earth elements is done in unsustainable ways and the processing of these materials can cause environmental degradation. From deforestation to soil erosion and water contamination.
Is there an environmentally friendly way to use AI?
The use of AI presents a double-edged sword: AI can drive positive change, but it’s a technology that’s extremely resource-heavy. Knowing if there’s an environmentally friendly way to use AI is a complex topic that’s in discussion.
Big Tech’s pledge to sustainable AI
We often report on Big Tech’s privacy washing strategies, and now something similar is happening, but this time the new buzzword they are marketing is sustainable AI.
Tech giants like Meta, Google, Amazon, and Microsoft are rapidly developing AI models and integrating them into their products as much as possible. But of course, they also know there’s a massive environmental impact which is why they often promise carbon neutrality and to put recycling measures in place. Similar to privacy washing, where they appear to care for user privacy, it is common for them to market sustainable AI to be perceived as environmentally conscious.
How environmentally conscious these tech giants are is still unknown, but there are reports of their increased carbon emissions. According to a report by the United Nations, the indirect carbon emissions from Meta, Amazon, Microsoft, and Google rose by an average of 150% from 2020 to 2023 due to the high energy demands of their data centers.
Another interesting finding is that Google who set the goal of being net-zero in carbon emissions by 2030, quietly removed the net-zero carbon goal from its website. While the net zero goal is apparently still in the company policy, given it’s no longer advertising it front and center and the fact it’s heavily investing in AI, it’s not convincing that this goal will be reached.
What can you do?
The boom in AI, specifically generative AI, has the potential to do great things, but it also has the potential to cause a lot of harm. Not only to the environment, but to us as a society too, for example, through giving biased results. While it’s not clear how bad AI is for the environment, there’s enough evidence to know it is not good for it. So the next time you open ChatGPT or your AI of choice, think twice before asking questions, to write your emails, or run tasks, because in the background the environment is paying a heavy cost for your convenience.
