Skip to content

There seems to be a shiny new AI tool for just about everything at the moment. From chatbots to virtual assistants to content creators of all shapes and sizes, each one attempting to revolutionise some part of our personal and working lives.

While these advancements are undoubtedly impressive and sometimes hugely useful, they come with a hidden cost that’s worth looking more closely at: the environmental impact.

In this article, we’ll look at the hidden environmental impacts of AI and ask whether we, as businesses and as individuals, can make use of these tools while limiting the cost to the planet…

The rapid growth of AI

The development of mainstream AI tools has gathered significant pace over the past few years. What was once the domain of science fiction has quickly become an integral part of our lives, and it’s doing some incredible things.

AI can translate languages in real-time. It can transcribe lengthy videos and auto-generate captions to help with accessibility. It can write emails and organise meetings. It can predict sales and order the right amount of stock. And now it can even generate creative content like text, imagery, code and even music. The capabilities of AI are genuinely staggering.

But these tools don’t come without a cost. And while we are all aware of the ethical and moral dilemmas associated with AI’s capabilities, its rapid growth and widespread adoption comes with a significant environmental footprint that is often overlooked.

As the development of AI tools has expanded and picked up pace over the past few years, the energy being used to train the models and power the systems behind them has skyrocketed.

The use-cases of AI

AI tools and applications span a wide range of uses. Some of these can be seen as being profoundly beneficial on both an individual and a humanity-wide scale. Others arguably less so.

Here are a few examples that span the spectrum of what we might call ‘essential’:

  • Healthcare: AI systems can be trained to support with medical diagnoses, for example by analysing images and other medical data to help detect the early signs of cancer and other diseases. This means treatment can be given earlier, improving the chances of better recovery and outcomes.
  • Climate modelling: Ironically, given the topic we’re looking at here, AI can be used to help predict climate patterns and assess and project the impact of various environmental policies. It could genuinely help us in the fight against climate change.
  • Customer service: Very few people want to have to pick up the phone to customer services these days, and AI has been a massive help for businesses and consumers alike by enabling self-service through chatbots and automated messaging. Not only does this mean people can handle their own queries and issues in their own time, it frees up actual humans for more complex tasks.
  • Social media algorithms: AI is used by social media platforms to curate content, ostensibly to be helpful, but essentially to keep users engaged (addicted) and on their site. To put it lightly, this has had a mixed social impact, and relies on users giving up their data privacy. Which is not cool.
  • Luxury applications: AI-powered personal shoppers! Virtual influencers! These may be helpful for individuals, but how essential are they really?

Let’s look at some stats on the impact of AI, then you can decide if the convenience is worth the cost…

Statistics on the environmental impact of AI

The leading AI companies are tight-lipped about the exact figures around the environmental impact of their creations, but there are a few key statistics that can provide a bit of insight into the potential scope and magnitude of the issue. For example:

  • Water consumption: At Microsoft’s Iowa data centres, water usage tripled in 2022 during the training of ChatGPT-4, coinciding with a severe three-year drought in the area.
  • Water use: Researchers estimate that, for every 20 questions answered, ChatGPT-4 consumes up to 500ml of water.
  • Energy consumption: Training the earlier and smaller GPT-3, with 175 billion parameters, emitted the same amount of CO2 (550 tonnes) as a single person taking 550 roundtrip flights between New York and San Francisco.
  • Scale of use: A study by European law enforcement agency Europol estimates that, by 2026, as much as 90% of online content will be generated or manipulated by AI.

How does AI impact the environment?

The environmental impact of AI is hard to quantify because it is so multifaceted. It’s not just about the massive amount of energy required to train AI models and the resulting carbon emissions we need to consider; there’s also the amount of water being consumed by data centre cooling systems, and the impact of mining for the precious metals and other raw materials required to build the hardware.

The extraction, processing and transportation of these metals is contributing to further ecological destruction and pollution in areas that have often been exploited for years already.

If the data centres being used to power the AI aren’t using green energy, that means fossil fuels are being used – and the ecological and environmental impact of that is well-documented.

Essentially, building and training AI is taking an awful lot from the planet, and it isn’t necessarily giving back.

The AI efficiency dilemma

The paradox, of course, is that this technology can be an incredibly powerful tool in the fight against climate destruction. Alongside its capabilities in supporting with climate modelling, AI can help reduce energy usage and carbon emissions by increasing efficiency within IT systems and business processes. For example by automatically switching between data centers based on the realtime conditions of each location.

But it’s not as straightforward as that.

For one thing, if businesses can save time and reduce waste by optimising processes (such as with those self-service customer chatbots we’ve spoken about), they are more likely to use AI tools frequently, thereby increasing the amount of energy and resources required to run the AI systems.

Moreover, how many businesses, when saving time and gaining what’s essentially extra hours in the day, will switch off and go home rather than simply do more work? If a business can theoretically make itself twice as efficient, will it be happy to clock off at lunchtime and use less energy? Or will it instead do twice as much work in an attempt to make double the money?

This is something sustainable web expert Claire Thornewill addresses in this excellent LinkedIn post, where she asks whether the much-trumpeted ‘increased efficiency’ brought about by AI tools will simply just give us more time to create more pollution. This is an important consideration for any business looking to adopt AI technologies, but especially for those that want to operate in a sustainable way.

How can we use AI sustainably?

For small businesses in particular, AI represents a powerful (and persuasive) opportunity to compete with those that have bigger teams and bigger budgets. On things like customer service and marketing, AI tools offer a genuine option when there just aren’t the resources to fund an extra person’s salary.

But AI has an environmental impact that’s often hidden from those making the decision to use it. And as we inevitably continue to integrate these tools into our lives and businesses, it’s crucial to weigh the environmental cost against any potential gain.

Here are a few things we can do to minimise the environmental impact of AI:

  • Evaluate necessity: Before turning to AI for a task, consider whether it’s truly necessary or if you can complete the task without it.
  • Optimise the use of AI tools: If you’re using AI tools, make sure you’re using them mindfully, and in a way that is optimised for energy efficiency. For example, don’t have an AI writing tool plugging away on social media posts every day if you only have time to publish and engage three times a week.
  • Block AI crawlers: Consider blocking AI crawlers from scraping your website for content to use in its ‘training’. Not only will this reduce unnecessary energy use associated with your site, it will prevent unnecessary data processing.

Conclusion

The advancements in AI over recent years undoubtedly represent an opportunity for businesses and individuals to save time and money while expanding their horizons. But the environmental cost of using these powerful tools cannot go unchecked.

In this article, we have looked at how:

  • Training AI models uses huge amounts of energy, generating an estimated 5x the lifetime carbon emissions of the average car.
  • AI has become a central feature of everyday life for many people, with ChatGPT receiving more than 1.5 billion visits in March 2023 alone.
  • If AI use continues to grow at its current rate, by 2025, AI tools could consume more energy than the entire global human workforce.
  • Saving time by using AI won’t necessarily lead to less work – in a capitalist economy, time is money, and it’s more likely to drive the call for more work to create more wealth (and more pollution).
  • We can potentially use AI sustainably within our businesses if we are mindful about where, when and what we use it for.

By arming ourselves with knowledge about the environmental impact of AI and making mindful and sustainable choices about how we use it, we can harness these powerful tools in a way that benefits both our businesses and the planet.

Further reading

beautiful websites,
rooted in good ethics

get in touch
Back to top Get in touch