AI and Sustainability

With sustainability becoming a pressing global issue, it is evermore vital for governments, corporations and organisations to come together to address this challenge. The rapid advancements in AI development could offer incredible tools and solutions to the fight for sustainability. AI excels at inference, the process of analysing vast amounts of data, identifying patterns and predicting outcomes, making it invaluable for solving problems when applied to sustainability issues and Earth sciences. Despite the immense promise that AI holds, its potential cannot be fully realised unless we address the underlying environmental impact of training and deploying these models.

AI has already been adopted by major corporations to enhance sustainability, this is particularly evident in production and distributions systems where AI is implemented to increase efficiency while decreasing energy consumption. Amazon’s Scout system optimises delivery routes by analysing customer preferences, traffic and weather conditions resulting in 10% reduction in carbon emissions, a 20% decrease in transportation costs and 30% cut in delivery time. Beyond logistics, AI’s ability to analyse images and uncover patterns has been utilised in examining satellite images, a much more efficient way as opposed to manual inspection. This process has helped oversee activities such as deforestation, monitoring ocean heath and assessing the impact of natural disasters. A study conducted in Dire Dawa, Ethiopia, demonstrated AI’s potential to identify flood-vulnerable zones in the region using data from historical flooding events in 2006 that affected under 4000 people. This technique enabled proactive measures to mitigate the impact of future disasters which is becoming increasingly critical as the severity of these natural disasters are estimated to increase in the face of climate change.

Despite these incredible advancements, the environmental costs of AI cannot be overlooked. AI development is energy-intensive with significant environmental costs occurring during two particular phases: training and deployment. Training AI models often takes weeks or months, requiring continuous fine-tuning, while deployment involves millions of daily user interactions all of which contribute to energy high energy consumption - 70% of which comes from fossil fuels, a non-renewable resource and high carbon emitter. Both processes rely on computers being housed in data centres that produce electronic waste containing hazardous substances like lead and mercury and exorbitant water consumption used for cooling systems. It is becoming increasingly alarming, the strain on local freshwater resources that these centres pose and even worse, exacerbating water scarcity in drought-prone regions like Arizona and Chile. It is predicted that AI could cause a rise in power demand up to 50% within the next 10 years in Europe alone, with global demands projected to rise tenfold by 2026.

In order to use AI while mitigating its environmental damage, it is essential that governments and private sectors to collaborate to ensure its sustainable development. The UN emphasises the need for standardised procedures to ensure AI’s environmental impact, as current data on this issue is limited and thus unable to spread awareness. Tools such as Carbontracker, proposed by L.F.W Anthony, can be used to predict the energy consumption, duration and carbon footprint of training a deep learning model. The predictions that this system can make supports transparency by encouraging companies to disclose their environmental consequences and can further stop model training if the predicted cost is exceeded. Secondly, the UN emphasises the importance of tech companies improving AI algorithms to be more efficient to reduce energy demands and recycling water and components where possible. The University of Queen Mary exemplifies this through their innovative way of recycling waste heat produced by their computer systems help heat campus buildings.

In conclusion, while AI is at the forefront off innovation and groundbreaking opportunities to address sustainability challenges, it is imperative to balance its benefits with responsible development. This can be achieved if the government and big tech corporations work together to ensure a sustainable future through innovation and transparency. With the right amounts of effort, AI can become a powerful tool in sustainable development and the preservation of our planet.