AI's paradox: Climate solutions and its ecological cost

Artificial Intelligence often conjures images of a futuristic world filled with smart devices, automation and predictability. However, like every major technological revolution, AI brings along its own set of repercussions. Remember the combustion engine? Once celebrated for its power, it's now at the heart of the climate debate, particularly in its contribution to environmental degradation. Earlier this year, the EU finally decided that after 2035, no new cars may be registered in EU states that run on diesel or gasoline. Similarly, AI, despite its wonders, is causing concern due to its significant ecological footprint.

It might surprise many to learn that some consider the AI industry an "extractive" business. Scholar Kate Crawford offers an eye-opening perspective: AI systems have a voracious appetite, not only for energy but also for rare earth minerals. This hunger translates into a considerable carbon footprint, comparable to the energy consumption of entire towns. With AI's capabilities to understand, predict, and even replicate human behaviours, the question arises: Can't it be more environmentally conscious?

Training an AI model is a resource-intensive task. Companies are often tight-lipped about their energy consumption, but diligent researchers make estimations based on the data they can gather. A study by Shaolei Ren from the University of California found that training GPT-3, the engine behind ChatGPT, may have consumed up to 700,000 litres of freshwater. This water, essential for cooling data centres, unfortunately evaporates post-use, rendering it non-reusable. Moreover, the process of obtaining raw materials for AI hardware often involves mining, which has both environmental and labour concerns.

According to the US Department of Energy, data centres are incredibly energy-intensive, using up to 50 times the energy per floor space compared to a typical commercial office building. This accounts for nearly two per cent of the total US electricity consumption.

However, it's not all doom and gloom. AI holds significant promise in addressing climate change. To understand this, it's essential to grasp how AI functions. At its core, AI algorithms learn and improve by processing vast amounts of data. As more data is processed, the predictions become more accurate. Such enhanced predictions can greatly assist governments in managing environmental challenges like flash floods, droughts, or unexpected climate shifts. Traditional systems struggle to predict such rapidly changing events, a gap AI can fill.

In sectors like energy, AI's transformative potential is evident. We no longer have to depend on rough estimates for electricity demand. AI allows for dynamic models that adapt and learn, paving the way for reduced reliance on carbon-intensive power sources. We're transitioning from mere production to intelligent allocation.

AI is playing a role in reducing pesticide use through precision agriculture, optimising fuel consumption by identifying and correcting inefficient driving habits, and predicting electricity demand more accurately. Moreover, as urban areas evolve, it becomes pivotal in advancing waste management. AI-based technologies like intelligent garbage bins, classification robots, predictive models, and wireless detection enable the monitoring of waste bins, predict waste collection, and optimise the performance of waste processing facilities. Better segregation of waste is leading to a global revolution in generating renewable energy.

Major corporations are recognising AI's environmental potential as they steer ahead on their ESG and sustainability journey. AI-driven maintenance prevents energy wastage, reduces risks, and enhances operational efficiency. Coupled with the Industrial Internet of Things, industries like aerospace are experiencing improved product designs, safety, and overall efficiency. Tech giant Google’s "4Ms" framework offers a responsible approach to AI, focusing on efficient models, AI-optimised processors, cloud computing, and strategic data centre placement.