The Invisible Monster: The Carbon Footprint of AI
summary
This talk explains the carbon footprint of generative AI and large language models, quantifying emissions from training (e.g., GPT‑3.5) and from inference/usage such as coding assistants like GitHub Copilot or CodeWhisperer. It compares the impacts of training, fine‑tuning and serving models, and demonstrates concrete tools and GreenOps hosting practices to measure and reduce AI-related CO2 emissions.
AI-written summary · about
