Meta engineers Lisa Rivalin and Richa Mishra explain how help the company manage power consumption, hardware reliability and carbon footprint across its fleet. The discussion covers specialized agents with human oversight, utilization tracking, early warning systems for hardware failures, and carbon accounting methods.
The session also explores how Meta extends hardware lifespan through component reuse and collaborates with the industry on more sustainable AI infrastructure. Topics include the relationship between AI growth, supply chains and environmental impact, and why reliable code and human approval matter when deploying AI agents at scale.
This is an AI-written summary. The original reporting belongs to Imagination in Action. How we write these · Spot a mistake?