The Environmental Impact of AI
AI is a hot topic. It offers huge benefits, with potential time and resource savings in all business sectors … and increasingly our everyday lives. A key question remains: AI is super-efficient, but is it environmentally friendly?
This is important because Artificial Intelligence (AI) usage and development is expanding … very fast. Since 2012, the amount of computing power used for deep learning research has doubled every 3.4 months. This equates to an estimated 300,000-fold increase from 2012 – 2018. (Scientific Computing World). As demand grows, so does AI’s energy consumption, increasing carbon costs. Machine learning and deep learning applications consume an enormous amount of energy; the energy consumption of data centres is about 40 times that of traditional office buildings. (Science Direct)
Information and Communications Technology (ICT) already accounts for approximately 4% of worldwide carbon emissions. Its contribution to greenhouse gas emissions is 60% higher than the aviation industry’s 2.4%. (The Shift Project research, Our World in Data)
A senior researcher at Huawei predicts that by 2025, data centres will account for 33% of global ICT electricity consumption. Organisations interested in sustainability should investigate options to reduce their AI’s carbon footprint.
Why is AI so power-hungry?
Two key activities drive AI’s energy consumption:
- Learning – a system developed by the University of Massachusetts analysed the carbon cost of machine translation and language modelling involved with an AI model. The result was comparable to the amount produced by 125 round trips from New York to Beijing!
- Availability – data centre location and operation impact hugely upon energy efficiency. The key issue is cooling. In a typical data centre, 30% of electricity is used by functional devices. 45% is used by thermal management systems including air conditioning, dehumidifying and chilling (Science Direct).
Microsoft is investing £2.5bn to expand AI data centres across the UK. Google is investing £790m to build its first UK centre; to be located in Hertfordshire and completed by 2025. Google’s greenhouse emissions increased by 48% between 2019 and 2023, primarily due to increases in data centre energy consumption and supply chain emissions. (BBC) The new facility plans to use the heat generated by its new data centre to heat homes and businesses in the local area.
As with all solutions, there are choices. “Green” AI enables your organisation to:
- offer sustainable workplace efficiency, and
- deliver eco-friendly practices.
Optimised airflow systems and innovative rack-level cooling reduce energy consumption and enhance overall performance. These methods save up to 67.2 % of energy compared to traditional methods. (International Journal of Refrigeration) Surely more data centres should consider these practices?
How can AI become “green”?
Partnerships and transparency are key to creating more energy-efficient AI solutions:- Optimal operational efficiency can be achieved via energy-efficient algorithms and equipment.
- Proactive data management, including minimising data accumulation, reduces energy consumption.
- Accountability can be achieved by disclosing and monitoring the environmental impact of AI.
- Strict guidelines regarding the ethical development, and management of AI technologies should be enforced by governments and regulatory bodies.
‘Most people are amazed by how much energy using AI involves,’ says Eric Hughes of EMH Technology in Hitchin. ‘Undoubtedly its application and usage are growing massively, but so is its impact on our environment. There are opportunities for providers to increase efficiencies. This will appeal to many organisations and individuals keen to minimise their own environmental impact.’
How does your organisation use AI?
Have you considered the environmental impact of the options available?
Explore IT management solutions with the experts at EMH Technology. Enjoy an initial discussion without obligation or jargon. We love discussing the potential of IT for organisations.