Lecture Notes on AI's Water Consumption

Introduction

• Presenter: Andrew Chang
• Topic: The hidden cost of AI related to water consumption.
• Illustration: Drinking a bottle of water to symbolize the water AI uses to power and cool supercomputers.

Key Points

AI Water Consumption

• Study Insight: Each conversation with Chat GPT represents significant water consumption.
• Scale: 100 million active users; each conversation adds to the total consumption.
• Perspective: Water usage often overshadowed by focus on greenhouse gas emissions and electricity use.

General Water Usage

• Average Usage: Canadians use around 215 liters/day (flushing toilets, cleaning, etc.).
• Industrial Use: Accounts for 5-10% of the world's freshwater (textiles, chemicals, paper, energy production).
• Agriculture: Utilizes 70% of freshwater for food production.

Data Centers and Cooling

• Importance of Cooling: Supercomputers, like those powering AI, generate substantial heat.
• Water as a Cooling Agent: Essential due to high thermal capacity; more efficient than air.
• Example: Microsoft’s supercomputer specifications to train AI (10,000 GPUs, 285,000 CPUs).

Water Treatment and Cooling Systems

• RC Harris Water Treatment Plant: Produces 30% of Toronto's drinking water.
• Cold Water: Essential for cooling data centers.

Case Study: Toronto Metropolitan University

• Chiller System: 4,000-ton chiller plant cooling 2.9 million sqft; uses a closed and open loop system.
• Evaporative Cooling: Similar to human sweating, losing water to cool a large area.
• Water Loss: System loses the equivalent of 8 Olympic-sized pools of water in summer.

Freshwater Scarcity

• Global Water Issues: Over a billion lack access to clean water; rivers and lakes drying due to human activity.
• Future Predictions: By 2025, two-thirds may face water shortages.
• AI and Other Technologies: Increasing demand for water as they grow.

Expert Insight

• Interview with Brad Bass: University of Toronto expert on green technology and water.
• Understanding Water Usage: Emphasizes unseen water usage in daily life.
• Concerns: Global supply and impact of technology on water bodies, especially in areas without large water bodies.

Conclusion

• AI's Water Footprint: While small compared to total human use, its growth could have major implications.
• Awareness: AI's widespread application could impact water resources.
• Final Thought: AI, like humans, requires water—a reminder for future technological advancements.

These notes provide a high-level overview of the lecture content focused on the environmental impact of AI technology, specifically its water consumption.