The main points of an environmental-aware AI approach, sometimes called frugal AI, include the following:

• The strategy involves assessing the environmental impact and sustainability of AI model training and management activities. This includes evaluating energy consumption, water consumption, and greenhouse gas emissions associated with large-scale, high-performance computational resources used for AI systems.
• Key indicators of AI system energy and water consumption, efficiency, and GHG emissions are identified and implemented. Measurable baselines for sustainable AI system operation are established in accordance with organizational policies, regulatory compliance, legal frameworks, and environmental protection and sustainability norms.
• Environmental impact indicators are included in both AI system design and development plans. This involves reducing consumption and improving efficiencies to minimize negative environmental impacts.

These points collectively form an environmental-aware AI strategy, focusing on minimizing negative environmental impacts while leveraging AI’s potential to address societal and environmental challenges.