AI already uses as much energy as a small country. It’s only the beginning.

🌈 Abstract

The article discusses the growing energy demands of artificial intelligence (AI) and data centers, and the need for more transparency and greener practices in this area.

🙋 Q&A

[01] The Rise of AI and Energy Consumption

1. What is artificial intelligence, and why is it getting so much attention?

• Artificial intelligence has been around since the 1950s, and has gone through cycles of hype and disappointment ("AI winters" and "AI summers").
• The current "AI summer" is focused on generative AI, which should be critically evaluated to determine if it should be used in applications where it wasn't used before.

2. What do we know about the energy costs of the current "AI summer"?

• It's difficult to measure the exact energy usage of AI, as it is a distributed and opaque process.
• Research has shown that switching from a non-generative AI approach to a generative one can increase energy usage by 30-40 times for the same task.
• The energy and environmental impacts are adding up as more data is stored, models are trained, and AI systems are deployed.

[02] The Environmental Impact of AI and Data Centers

1. What are the material and environmental impacts of supporting AI and data centers?

• Connected data storage (e.g. cloud storage) consumes more energy than static storage.
• Training AI models requires a lot of energy, with thousands of chips running for thousands of hours.
• Newer generations of AI hardware (e.g. GPUs) are more powerful but also more energy-intensive.
• Data centers need a lot of energy to run, often relying on carbon-intensive energy sources like natural gas and coal.
• Data centers also require significant cooling, which uses a lot of water.

2. What practices and policies should be considered to address the environmental impact of AI?

• Providing information and ratings (e.g. an "Energy Star" rating) to allow consumers to choose more energy-efficient AI models.
• Promoting "digital sobriety" - encouraging consumers to critically evaluate their need for new AI-powered gadgets and services.
• Ensuring data centers are built in locations with access to renewable energy sources.