These past few years, AI has taken the world by storm. Nearly transforming every walk of life, the wide-ranging tool has changed how we think, use data, and operate in the sea of information. Wish to hear your favourite artist sing a song they’ve never performed? No problem. Stuck in a writer’s block for your essay? ChatGPT will take care of it.
The opportunities for using AI are endless, and people are taking advantage of it. In fact, the data from August 2024 suggests that ChatGPT has around 200 million weekly active users worldwide. But what does it mean for the environment? Do AI’s benefits outweigh the negative consequences? Can AI be used to improve sustainability? Let’s find out!
Measuring AI’s environmental impact
Figuring out the exact impact AI has on climate is still quite tricky. Most research focuses on the amount of greenhouse gases emitted; however, this is challenging. What makes it so difficult is that different types of AI require various quantities of computing power to operate. For example, Large Language Models (LLMs), such as OpenAI’s GPT-3, are more damaging as they require an astonishing 500 tons of carbon dioxide to be trained. This is the equivalent of 123 gasoline-powered passenger vehicles.
However, many simpler AI models aren’t as harmful to the environment, as they don’t produce as many emissions. The issue with calculating the effect of AI on the climate crisis is that there is a lack of transparency in many AI companies. Nevertheless, most experts agree that AI has the potential to both hurt and help the environment, depending on the way it is used.
Our Tip
If you’re using AI tools like ChatGPT frequently, try to consolidate your tasks into fewer sessions. By doing so, you’ll indirectly save energy and reduce the carbon footprint of the AI systems powering these tools.
The challenges of AI for the environment
The biggest criticism of AI systems and their data centres is related to the amount of energy needed to power them. In most cases, they require much higher energy consumption than traditional IT and cloud solutions. Moreover, most of the data centres get their energy from burning fossil fuels, which creates greenhouse gases.
While Google Search has often been scrutinised for its energy consumption, ChatGPT takes it to a whole new level. In fact, making a single request to its AI-based system uses 10 times the electricity. Almost all best-known generative AI models need hyperscale cloud providers with an enormous number of servers. The Financial Times, for example, has reported that AI data centres in the U.S. demand 15 gigawatts of power annually – the amount that all U.S. solar farms use in a year.
But it’s not only about energy. Electronics used in these sophisticated data centres require a staggering amount of materials. In fact, to make a 2 kg computer, around 800 kg of materials will be needed. Shocking, isn’t it? Likewise, the microchips used for powering AI usually include rare earth elements, which are destructive to the environment in their own ways.
The third issue highlighted in regard to AI and sustainability is electronic waste. Unfortunately, these data centres often produce heavy waste, including hazardous substances like lead and mercury.
Experts are also raising concerns about the enormous water consumption at these centres. Some estimates suggest that AI systems might soon consume six times more water than Denmark. This, of course, is worrisome, especially when some parts of the globe still don’t have access to clean water.
Did You Know
Training a single AI model or chatbot can use more electricity than 100 homes in a year.
Bloomberg UK
So, is it all bad?
Does AI have a higher carbon footprint? Yes, it does. But it’s not all negative. In many ways, the use of AI can contribute to sustainability. Over these years, we’ve seen how it can optimise supply chains, help manage energy consumption, and even enable energy storage, which, in the end, reduces the number of total emissions. The real question is striking a balance between the technology’s sustainability benefits and energy demands.
The good news is that many data centres are investigating their energy efficiency, which should help mitigate their carbon footprint. With an expected annual growth rate of 46.% between 2024 and 2030, it’s clear that AI is here to stay, and more industries will rely on it. So, the question is – how can we take advantage of the enormous potential AI has while not hurting the planet along the way?
Ways to improve AI’s sustainability
As of this moment, more than 190 countries are following recommendations on the ethical use of AI. This is a good step forward. However, there’s still a lack of environmental guidelines that could help reduce the negative impact of AI.
With AI’s projected growth, the need to cut emissions and choose facilities that operate on renewable energy is more critical than ever. Thankfully, tech giants are paying attention to this, too. Microsoft, for example, is trying to add primarily renewable energy to the grid and has been investing in carbon sequestration.
Furthermore, many companies are now looking into optimising their algorithms and making them more energy efficient. The goal is to reduce model sizes (effective models can be developed without extensive data) and produce less energy-demanding GPUs. Another solution is to focus on prompt engineering and tuning, which can help optimise hardware usage, reducing some of the carbon footprint.
Final thoughts
If we imagine a scenario where AI is used by billions worldwide, the total annual carbon footprint could be 47 million tons of carbon dioxide, which would result in a 0.12% increase in total emissions. There is no question that setting up frameworks for making AI sustainable is now more critical than ever.
While the challenges of using AI are real, it’s important to understand that AI also has the potential to be a transformative agent. When used correctly, it can be a tool to improve sustainability, not the other way around.
Interested to learn more about AI and green computing? Make sure to follow our blog for more useful guides and industry insights.
