Environmental Impacts of Artificial Intelligence (AI)

The discussion around AI primarily focuses on its benefits in sectors like healthcare and agriculture, but its environmental impacts are often overlooked.

Carbon Footprint of AI

• AI algorithms development increases carbon emissions, contributing to climate change.
• The global ICT industry's GHG emissions range between 1.8%-2.8% to as high as 2.1%-3.9%.
• A Google report claims low electricity usage for a single AI text prompt (0.24 watt-hours), but this is criticized as misleading.

Water and Energy Consumption

• UNEP's report projects AI servers could use 4.2 to 6.6 billion cubic meters of water by 2027, risking water scarcity.
• Training a single Large Language Model (LLM) may produce nearly 300,000 kg of carbon emissions.
• Energy consumption for AI models like ChatGPT is significantly higher than other digital services like Google search.

Global Legislative Efforts

• UNESCO highlighted AI's negative societal and environmental impacts.
• The U.S. and EU have proposed legislation for AI's environmental impact, such as the AI Environmental Impacts Act of 2024.

India's Response and Recommendations

• India should recognize AI's environmental costs and consider Environmental Impact Assessments (EIA) for AI developments.
• Stakeholders like tech companies and NGOs should collaborate to create standards for evaluating AI's environmental impact.

Data Collection and Reporting

• Employ sustainability metrics to assess AI's consumption of GHGs, energy, and water.
• Consider AI's environmental impact in ESG disclosure standards, drawing from EU's CSRD framework.

Sustainable AI Practices

• Adopt practices like using pre-trained models, renewable energy for data centers, and AI-specific reporting.

Author: Amar Patnaik, a lawyer and founding partner of A&N Legal Solutions LLP, former CAG bureaucrat, and ex-MP in Rajya Sabha.