Towards Environmentally Equitable AI

TL;DR

This paper tackles regional environmental inequity in globally deployed AI systems by proposing equity-aware geographical load balancing (GLB) to distribute environmental costs more fairly across regions. It formalizes the approach with a minimax fairness objective that couples traditional GLB costs with an equity regularizer: $\sum_{t=1}^T\sum_{i} cost_{i,t}(x_{i,t}) + \lambda\cdot\max_{i}\left[\sum_{t=1}^T E_{i,t}(x_{i,t})\right]$. The authors discuss algorithmic challenges, including unknown horizon terms, forecast uncertainty, and the tradeoffs between online guarantees and data-driven optimization, and propose learning-augmented online strategies to mitigate inequity. Finally, they outline directions such as coordinated training/inference scheduling, joint IT/non-IT optimization, holistic knob control, and theoretical foundations to enable environmentally equitable AI with acceptable total cost.

Abstract

The skyrocketing demand for artificial intelligence (AI) has created an enormous appetite for globally deployed power-hungry servers. As a result, the environmental footprint of AI systems has come under increasing scrutiny. More crucially, the current way that we exploit AI workloads' flexibility and manage AI systems can lead to wildly different environmental impacts across locations, increasingly raising environmental inequity concerns and creating unintended sociotechnical consequences. In this paper, we advocate environmental equity as a priority for the management of future AI systems, advancing the boundaries of existing resource management for sustainable AI and also adding a unique dimension to AI fairness. Concretely, we uncover the potential of equity-aware geographical load balancing to fairly re-distribute the environmental cost across different regions, followed by algorithmic challenges. We conclude by discussing a few future directions to exploit the full potential of system management approaches to mitigate AI's environmental inequity.

Figures (1)

• Figure 1: Comparison of GLB algorithms in terms of the total energy cost and the normalized water/carbon peak-to-average ratio (PAR). Details in Shaolei_Equity_GLB_Environmental_AI_eEnergy_2024.