AI is already being used to monitor methane emissions, track harmful sand dredging, and even predict future environmental outcomes by analysing data patterns. This technology could support governments and organizations in making eco-friendly decisions, giving hope for solutions to climate change, biodiversity loss, and pollution.
Is the environment ready for the impact of AI? The Environmental Cost of AI
1. Resource-Intensive Data Centres
AI needs vast data centres, which require enormous amounts of water, electricity, and minerals. The construction and operation of these facilities are resource-heavy; for instance, building a 2 kg computer uses 800 kg of raw materials.
2. Electronic Waste and Hazardous Chemicals
Data centres produce a lot of electronic waste, often containing toxic substances like mercury and lead, which harm the environment when improperly disposed of.
The Unpredictable Impact of AI
AI’s environmental impact can also be hard to predict, making it something of a wildcard. For example, AI-powered self-driving cars could encourage more car use, increasing emissions. AI could also be misused, such as in spreading misinformation about climate change, misleading public opinion on environmental risks.
UNEP’s Recommendations for Sustainable AI
To mitigate AI’s impact, the United Nations Environment Programme (UNEP) suggests five key actions:
- Standardized Measurements
Develop standard methods to measure AI’s environmental impact for more consistent data. - Transparency in Environmental Impact
Require companies to disclose AI’s direct environmental consequences to promote accountability. - Efficient AI Algorithms
Tech companies should make AI algorithms more energy-efficient and recycle water and components wherever possible. - Greener Data Centres
Encourage the use of renewable energy in data centres and carbon offset programs to lessen AI’s carbon footprint. - Integrating AI with Environmental Policy
Align AI policies with broader environmental regulations to create a sustainable framework for AI deployment.