AI 回彈效應抵消效率改善
Operation & Monitoring
Risk Description
While AI technology can improve energy efficiency and resource utilization, cost reductions from efficiency gains stimulate more usage, ultimately increasing total resource consumption instead of decreasing it (Jevons Paradox). For example, more efficient inference techniques lower per-query costs but promote AI application in more scenarios, actually increasing total energy consumption. The "more efficient, more used" rebound effect may undermine AIs green promises.
Framework Mappings
iso 23894R8
tw principle永續發展與福祉
tw risk type社會系統性影響(Societal Systemic)
Controls English translation pending
- 總量管控
- 使用合理性評估
Implementation Steps English translation pending
- 設定 AI 使用的能源消耗總量上限,而非僅追求單位效率
- 在部署新 AI 應用前評估其使用量增長對總體能耗的影響
- 建立 AI 碳預算制度,追蹤和限制組織的 AI 環境足跡
- 定期報告 AI 使用的總能耗和碳排放趨勢