Whether you love it, hate it, or have been taught by many movies to fear it, there’s no denying AI’s prominence in the world of tech. Artificial intelligence is being used across a huge number of industries to streamline processes, reduce admin time and just generally make our lives easier.
But with all new tech advancements, there’s usually a downside, and for AI (putting the Skynet associations to one side), that might be just how much energy it uses. Here we’re taking a look at the environmental impact of AI to see how much we should be using it while staying kind to the planet.
What is the Environmental Impact of AI?
AI is used for so much these days, from simple product recommendations to fully automated machinery. More people than ever are also using generative tools such as ChatGPT to create content, but this widespread use of AI could be doing more harm to the environment than most people realise.
All tech has a carbon footprint, and ChatGPT is no different. Electricity is needed to power the servers and water is needed for data centre cooling (and sometimes electricity generation). This itself isn’t newsworthy, but the amount of water and power it uses is staggering.
ChatGPT uses over half a million kilowatts of electricity every day, and a single ChatGPT conversation uses about fifty centilitres of water, although this does vary significantly based on where the AI is trained and hosted. The training process for a single AI large language model can have a carbon footprint equivalent to 125 round-trip flights between New York and Beijing.
Some reports project that the global AI energy demand will exceed the annual electricity consumption of a small country like Belgium by 2026, and by 2050, two thirds of the global population is expected to suffer from water scarcity, which begs the question of whether we can really afford to keep using AI at the rate we currently are.
Weighing up AI’s Pros and Cons
All this isn’t to say that AI doesn’t have its place. It has already become undeniable that this tech can benefit us all greatly.
While AI like ChatGPT does use a huge amount of electricity, it can also save individual users from using hours upon hours of battery life creating content, which when multiplied by the number of users, quickly adds up to its own significant carbon footprint. As well as saving time, AI can provide inspiration and new ideas and cover more ground in the research phase. All this combined with human expertise can potentially result in a better end product, and who doesn’t want that?
Looking beyond just content creation, AI can also potentially be used to undo the environmental damage it causes (and the many years of damage we all caused before AI was even a thing). AI is used to assist various sustainability systems, such as crop yields, renewable energy, drought mitigation and more, and with the earth in the position it currently is, we need all the help and fresh ideas we can get to ensure our future.
We just need to make sure that the cost doesn’t outweigh the benefits.
The tech giants have a huge carbon footprint already, and the resource consumption of AI only adds to their environmental debt. We really can’t be condoning big tech companies further capitalising on practices that harm the planet to increase their bottom line.
There’s also a disparity across different regions, which is pretty telling. For example, Google’s data centre in Finland was opened on 97% carbon free energy in 2022, but this figure for their data centres in Asia are only between 4 and 18%, suggesting that public perception in each region might have more weight than the desire to do the right thing, and considering the size disparity between the two countries, the good is quickly dwarfed by the bad.
What’s Being Done About AI’s Carbon Footprint?
The good news is that data centre power and cooling technology is advancing, which will help to reduce the energy needed by AI models. Data centre operators, like lots of business owners, are under pressure to achieve net zero emissions, and an increased use of renewable energy will mean that the energy used is more likely to be offset and could have a huge impact on the carbon footprint of AI.
AI models themselves are also becoming more efficient, with better optimised algorithms that accelerate AI training and inference, along with techniques such as weight pruning, quantisation, and energy efficient GPUs meaning each AI interaction should use less energy.
While we can also be quick to vilify big tech companies, it’s important to remember that real people do work there, and they don’t necessarily share those same values. The Tech Workers Coalition, which includes employees of Amazon, Google, Microsoft, Meta and more, marched to demand that their employers reduce emissions to zero by 2030, to stop making deals with fossil fuel companies, and to stop funding climate change deniers.
The Verdict – Should We Keep Using AI?
There is undoubtedly a place for AI, but we shouldn’t rely on it, and we should be self-regulating how much we use it.
AI can dramatically increase output and, combined with human expertise, can help businesses and content creators deliver a better service, but all users of generative AI should be mindful of the environmental cost, and ensure that the benefits are outweighing this.
However, it is the tech giants behind these AI models that need to take responsibility for the shocking amount of energy they use. New tech, no matter how revolutionary it is perceived to be, should be for the betterment of the planet, not at the expense of it.
The development of smart grid design, low emission infrastructure and climate change prediction models can help us towards a greener future, but those with the power to make the most impactful change need to be held to account.
Here at Illicit, we specialise in sustainable web design. Our lightweight CMS and super efficient front ends reduce resource consumption and improve load speeds. So, if you are looking to reduce your carbon footprint with your new website, give us a shout.