LLM-dependent memory operations in agents and RAG consume orders of magnitude more energy than vector methods, and resource-constrained hardware pays higher energy for lower quality.
Triple Bottom Line or Trilemma? Global Tradeoffs Between Prosperity, Inequality, and the Environment
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CY 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Addressing the sustainable AI trilemma: a case study on LLM agents and RAG
LLM-dependent memory operations in agents and RAG consume orders of magnitude more energy than vector methods, and resource-constrained hardware pays higher energy for lower quality.