TED Radio Hour Our Tech has a Climate Problem: Here's how we solve it
15 snips
May 10, 2024 
AI researchers and climate researchers discuss the environmental impact of AI models, mining for essential minerals, reimagining transportation for sustainability, battery recycling, and using satellites and AI to monitor greenhouse gas emissions. They also explore the challenges of space debris collisions and harnessing AI for wind energy and electricity supply forecasting to mitigate climate change.
AI Snips
Chapters
Transcript
Episode notes
AI Energy Consumption Opacity
- Quantifying AI's energy use is difficult, especially for training large language models.
- Tech companies are becoming less transparent about their products' energy consumption.
Early AI Energy Estimates
- A 2019 study estimated training a large language model emitted as much carbon as five cars over their lifetimes.
- More recent estimates suggest even higher energy consumption, comparable to 130 homes' annual energy use.
Bloom's Environmental Impact
- Bloom, an open large language model, used a year's worth of energy for 30 homes during training, emitting 25 tons of CO2.
- Other models like GPT-3 emit much more carbon; model sizes have grown 2000 times in five years, increasing environmental costs.