AI is an ever-increasing presence in the world around us, with online applications from Quizlet to Google to the infamous Chat-GPT offering AI tutors, summaries, and study aids. While these programs seem to present an ideal time-efficient solution for the time-deprived student, concerns have been raised over the environmentally detrimental effects of developing and maintaining these AI programs.

The term ‘AI’, an initialism which stands for artificial intelligence, is a catch-all term for technologies designed to process large amounts of information and identify patterns in order to complete tasks that may mimic human thinking or conversation. Many praise AI’s seemingly unending applications in fields such as cloud computing, image processing, and cybersecurity, but others criticise its exploitative possibilities as it has the potential to devalue human art and music, and exacerbate biases. Another problem with AI, which often goes unnoticed, is its environmental impact.

AI technology’s development, maintenance and disposal generate a large carbon footprint through a staggering energy-intensive process, which only increases as AI models and datasets develop and become more complex. In September 2024, the UN recorded that the number of AI data centres has risen to 8 million, from 500,000 in 2012, and this number is expected to grow. Running an AI program, such as ChatGPT, consumes 10 times the electricity of a regular Google search, emitting hundreds of tons of carbon. This entails disastrous environmental complications, contributing to air pollution, thermal pollution in water bodies, and the general warming of the planet that is inducing destructive climate change and extreme weather globally.  The process of maintaining generative AI data centres is also water intensive, as it requires onsite server cooling and offsite electricity generation, putting a strain on freshwater reserves. A study conducted by researchers in the US suggested that globally the water demand for AI could be half of the entire United Kingdom’s demand by 2027. This desperately needs to be addressed as more and more websites and programs develop and use AI.

The resources required to build the technology also come at a cost to the planet, with cobalt, lithium, and tantalum often mined in environmentally destructive and exploitative ways, damaging local communities and the natural world around them. Such an intense process also produces large amounts of waste, outlining an immediate need to develop environmentally suitable waste management and recycling facilities to deal with this. A 2021 study by the Geneva Environment Network found that only 17.4% of E-waste was properly collected and recycled, with the majority left in landfills and dumps. If this continues, the World Economic Forum has estimated that by 2050, the total amount of electronic waste generated will surpass 120 million metric tonnes annually. Improper E-waste disposal can lead to hazardous chemicals such as lead, mercury, and cadmium contaminating soil and water supplies, ultimately presenting a risk to human life and the environment. Exposure to these chemicals causes a myriad of health complications including negative birth outcomes and respiratory issues. Those in developing countries are often hardest hit by these issues, as the majority of E-waste is shipped there, often illegally, to be processed away from richer countries.

Systems are being established to deal with E-waste, such as Egypt’s E-Tadweer app. This allows users to dispose of old electronics at specific locations dedicated to safe and environmental recycling, and exchange them for vouchers to purchase new products. The Restart Project in London is another initiative that trains individuals in electronic repairs, helping to reduce electronic waste and save money. 

However, AI can also be applied to tackle environmental problems. Its ability to analyse large data sets makes it an asset in hazard forecasting, with deep learning algorithms providing sensitive and specific early warning systems for extreme weather events. Researchers from the British Antarctic Survey and the Alan Turing Institute have developed a program called IceNet, which is a predictive tool that uses AI to predict sea ice change through data collected from satellite sensors. Google Flood Hub also uses AI models to issue flood alerts, sending 115 million notifications to 23 million people globally in 2021. Admittedly, however, while AI can be applied to support climate change prediction and risk mitigation, it paradoxically feeds into the very problem that it supposedly helps to reduce. 

More than 190 countries are calling for investment in more sustainable AI infrastructure, adopting a series of recommendations on the environmentally ethical use of AI. However, these are non-binding; inevitably, environmentally harmful models will continue to be used unless there are stronger regulations that enforce transparency and action to tackle the consequences of AI. Sustainable practice needs to be prioritised as an imperative as AI technology develops and expands.