Environmental Impact of Artificial Intelligence

Summary of Key Points on the Environmental Impact of Artificial Intelligence (AI)

Environmental Impact and Concerns

• OECD Findings: The development of AI applications is associated with climate change concerns due to their carbon footprint. The global Information and Communications Technology (ICT) industry contributes approximately 1.8% - 2.8% of global greenhouse gas (GHG) emissions, with some estimates as high as 2.1% - 3.9%.
• Water Usage: A UNEP report forecasts that house AI servers could consume between 4.2 billion cubic meters (bcm) and 6.6 bcm of water by 2027, raising concerns of water scarcity.
• Carbon Emissions from AI Models: Training a single Large Language Model (LLM) may generate about 300,000 kilograms of carbon emissions, equivalent to the lifetime emissions of five cars.

Specific Energy Consumption Indicators

• ChatGPT Energy Use: Requests through ChatGPT consume 10 times more energy than those conducted via a Google search.
• Energy and Policy Study: The training of a large AI model released in June 2019 suggested emissions of over 626,000 pounds of CO2.

International Frameworks and Government Responses

• UNESCO Recommendation (2021): Released a non-binding recommendation that emphasizes acknowledging the negative societal and environmental impacts of AI, adopted by approximately 190 countries.
• U.S. and EU Legislation: The U.S. has the Artificial Intelligence Environmental Impacts Act of 2024, and the EU is working on harmonized AI regulations.

India's Context

• EIA Notification, 2006: Environmental Impact Assessments (EIA) should encompass AI model development, emphasizing the need for broader assessments.
• Policy Recommendations:
  • Develop measuring standards for evaluating AI's environmental impacts.
  • Implement data collection metrics related to GHG emissions, energy consumption, and natural resource usage.
  • Explore inclusion of AI environmental impacts in ESG disclosures by the Ministry of Corporate Affairs and SEBI.

Good Practices and Future Directions

• Encourage adoption of sustainable AI practices such as:
  • Using pre-trained models to reduce energy costs.
  • Utilizing renewable energy sources to power data centers.
  • Regularly reporting on AI-specific environmental impacts.
• Strive for AI integration into support systems for global sustainability goals.

This summary encapsulates the critical discussions around the environmental impacts of AI, emphasizing the need for policy interventions and data-driven approaches to mitigate adverse effects, while also highlighting the responsibilities of stakeholders in adopting sustainable practices.