Situation
It seems that every other email, every article, every post these days mentions AI. Whether it’s extolling the promises that generative artificial intelligence holds for a brighter future; the fears that writers, PR practitioners and journalists harbour against the tool; ways to make best use of it; AI hacks to make life so much better; prompts to use to make the most of it; how it’s destroying the Earth; how it’s saving the world. You get the picture …
This article, yes, focusses on AI as well, but hopefully helps map the current state of AI and how it impacts our environment, as well as provides suggestions for responsible use and future regulations to look out for.
A brief history of generative AI
It might seem like AI has blown up in popularity and use in just the past few years. Which isn’t entirely wrong. In fact, AI and the mathematics behind it has been around for nearly a century. The Markov chain and Markov models have been used since 1906 to model the behaviour of random processes like next-word prediction tasks. In the 1930s and 40s the beginnings of basic techniques for machine learning came into play, with scientists only developing the computers that were powerful enough to use them in the 1970s. The first primitive generative AI was ELIZA, a text chatbot created by Joseph Weizenbaum in the 1960s.
But, what we’re really here to talk about are the tools that have cropped up in recent years.
In 2018, the first generative pre-trained transformer (GPT) was created and released by OpenAI. This is the basis of the tools we see around us today. DALL-E is also an important development, it was created in 2021 by OpenAI and generates images from text prompts.
These two breakthroughs have continued to develop in recent years, leading to AI tools popping up seemingly every week, from chatbots like ChatGPT (and its subsequent updates), Claude, Gemini, DeepSeek and Grok, notetakers like Fathom, OtterAI, grammar tools like Grammarly and Wordtune, image generation like Midjourney, DALL-E 3, voice and music generation like ElevenLabs, Mruf, Suno, Udio… and this list doesn’t even scratch the surface of what’s out there.
Now that generative AI is pretty much institutionalised in our day-to-day lives, even with a relatively short history, let’s look at the environmental impact of these systems, and why responsible and ethical use is imperative.
Environmental impact of AI
While every component within the generative AI supply chain of generative AI impacts the environment, let’s focus on four main parts: energy consumption, water usage, electronic waste and critical minerals.
What drives generative AI’s development is learning and training. This takes a lot of computing power, which uses quite a bit of energy. Although it’s difficult to pinpoint exact numbers of the impact, it being so soon to genAI’s implementation and use, globally, the electricity consumption of data centres rose to 460 terawatts in 2022, making data centres the 11th largest electricity consumer in the world. It ranks between the countries of Saudi Arabia and France (OECD figures). According to the IEA, data centres and data transmission networks are responsible for 1% of energy-related greenhouse gas emissions, and in 2022, the estimated global data centre electricity consumption was 1-1.3% of global electricity demand.
To put that energy consumption into context, it’s estimated that one query through ChatGPT uses 10 times the electricity of a Google Search.
To keep in mind, however, is the use of renewable energy sources at large data centres. In 2021, Apple, Google, Meta and Microsoft either purchased or generated enough renewable energy to match all of their energy consumption, with Amazon purchasing enough to match around 85% of its consumption.
Water consumption is also high as it’s used to cool data centres. Heat output is transferred to a cooling fluid – usually water – in air handling units. That’s aside from the water used during construction of the data centres and the electrical components. One study by scholars at US Riverside and UT Arlington estimated that the global AI demand on water is projected to be 4.2 to 6.6 billion cubic meters in 2027, or, half of the total annual water consumption of the UK.
Alongside energy consumption and water usage, critical materials and rare elements are used in the components that power all of this work, like nickel, lithium and cobalt, which are often not mined sustainably. Making a 2kg computer uses around 800 kg of raw materials. The electronic waste generated at the end of the computer’s life can be hazardous to the environment as it contains materials like mercury and lead.
Can AI help tackle environmental emergencies?
It’s not all bad news. AI does make a positive impact on the environment. It is used to analyse patterns and extract meaning from massive amounts of data. This can be used to predict future trends at an unprecedented rate. This is being actively used to tackle environmental emergencies.
UNEP uses AI to detect methane emissions from oil and gas install
ations. Researchers from the University of Leeds use AI to map icebergs in Antarctica, measuring changes 10,000 times faster than a human could. Space Intelligence uses machine learning to map deforestation, reviewing and measuring forest conservation and restoration initiatives. The UN uses AI technology in its IKI Project to help communities in Burundi, Chad and Sudan with climate risk planning. 
Clearly, AI is being used in many ways to help mitigate the effects of climate change and reverse damage. From waste recycling to ocean cleanup, predicting natural disasters to reforestation. Artificial intelligence is undoubtedly an asset when it comes to tackling environmental emergencies, assisting organisations in sifting through data in order to make informed decisions, spot patterns and identify anomalies, all at a rate impossible to achieve from human labour.
It would be irresponsible (if not impossible) at this point to cease all use of AI because of its negative environmental impact. But, the responsible use of AI is key here. When a single query on ChatGPT is similar to dumping out an entire bottle of water in terms of environmental impact, unfettered and unregulated use of this tool based on relentless demand cannot continue.
So, what are countries doing about it? (Regulations)
With such a new technology with rapid expansion and numerous impacts, whether environmental, economic, social, developmental, educational, there must be reams of regulations now to protect users and the planet, right?
Not really.
While each country has its own regulations that intrinsically have a
pplications to AI, the EU has the first international agreement on artificial intelligence, the AI Act. The purpose of this act was to help AI develop in Europe in a responsible way, with rules for developers based on posed risk (unacceptable, high, limited and low). This act has been recently updated, with the publication of a third draft of a code of practice. Globally, there are no other regulations like it.
The UK released its AI Regulation White Paper in 2023, communicating the opportunities and challenges that AI poses, as well as showing that the UK does not intend to enact AI regulations anytime soon. Instead, there are several offices (AI Policy Directorate in the DSIT, formerly the Office for Artificial Intelligence), strategies (National AI Strategy) and plans (AI Opportunities Action Plan) in place to help protect the development and use of the technology.
In comparison, the US has no federal legislation or regulations that hinder or accelerate the development of AI or its use. Instead, Americans rely on already existing federal laws, state regulations and guidelines. But, Congress is currently considering a number of bills on AI. For now, though, existing laws that have an application on AI are the sole regulatory basis for the US.
*For more information on regulations, I found this resource particularly useful.
Recommendations
Now that we’ve established that there are limited regulations in place, what are the recommendations of use?
UNEP recommends that:
- Countries can establish standardised procedures for measuring the environmental impact of AI
- Governments can develop regulations that require companies to disclose direct environmental consequences of AI-based products and services
- Tech companies can make AI algorithms more efficient, reducing energy demand, while recycling water and reusing components when feasible
- Countries can encourage companies to green their data centres (using renewable energy, offsetting carbon emissions)
- Countries can weave AI-related policies into broader government regulations
Given its immense environmental impact what can we personally and individually do to protect the environment, while also using this new piece of technology in helpful ways?
The most powerful tool when faced with dilemmas like choosing innovation over environment, or balancing both, is to get educated. Learn the different aspects of the situation, the data and statistics that accompany them and listen when leaders explain guidelines, their thoughts and how to move forward. We don’t really know how to solve this, or where these types of technology will be in a week, in a year, in a decade. So, I recommend staying up to date on the facts, and to think about limiting the queries on ChatGPT, for now.
The post AI and the environment appeared first on Resonates.
