Search for: All records

The enormous growth of AI computing has led to a surging demand for electricity. To stem the resulting energy cost and environmental impact, this paper explores opportunities enabled by the increasing hardware heterogeneity and introduces the concept of Geographical Server Relocation (GSR). Specifically, GSR physically balances the available AI servers across geographically distributed data centers subject to AI computing demand and power capacity constraints in each location. The key idea of GSR is to relocate older and less energy-efficient servers to regions with more renewables, better water efficiencies and/or lower electricity prices. Our case study demonstrates that, even with modest flexibility of relocation, GSR can substantially reduce the total operational environmental footprints and operation costs of AI computing. We conclude this paper by discussing major challenges of GSR, including service migration, software management, and algorithms. 

Public models offer predictions to a variety of downstream tasks and have played a crucial role in various AI applications, showcasing their proficiency in accurate predictions. However, the exclusive emphasis on prediction accuracy may not align with the diverse end objectives of downstream agents. Recognizing the public model's predictions as a service, we advocate for integrating the objectives of downstream agents into the optimization process. Concretely, to address performance disparities and foster fairness among heterogeneous agents in training, we propose a novel Equitable Objective. This objective, coupled with a policy gradient algorithm, is crafted to train the public model to produce a more equitable/uniform performance distribution across downstream agents, each with their unique concerns. Both theoretical analysis and empirical case studies have proven the effectiveness of our method in advancing performance equity across diverse downstream agents utilizing the public model for their decision-making. 

Fueled by the soaring popularity of foundation models, the accelerated growth of artificial intelligence (AI) models’ enormous environmental footprint has come under increased scrutiny. While many approaches have been proposed to make AI more energy-efficient and environmentally friendly, environmental inequity — the fact that AI’s environmental footprint can be disproportionately higher in certain regions than in others — has emerged, raising social-ecological justice concerns. This paper takes a first step toward addressing AI’s environmental inequity by fairly balancing its regional environmental impact. Concretely, we focus on the carbon and water footprints of AI model inference and propose equity-aware geographical load balancing (eGLB) to explicitly minimize AI’s highest environmental cost across all the regions. The consideration of environmental equity creates substantial algorithmic challenges as the optimal GLB decisions require complete offline information that is lacking practice. To address the challenges, we introduce auxiliary variables and optimize GLB decisions online based on dual mirror descent. In addition to analyzing the performance of eGLB theoretically, we run trace-based empirical simulations by considering a set of geographically distributed data centers that serve inference requests for a large language AI model. The results demonstrate that existing GLB approaches may amplify environmental inequity while eGLB can significantly reduce the regional disparity in terms of carbon and water footprints. 

ABSTRACT Betacoronavirusesencode a conserved accessory gene within the +1 open reading frame (ORF) of nucleocapsid called the internal N gene. This gene is referred to as “I” for mouse hepatitis virus (MHV), ORF9b for severe acute respiratory CoV (SARS-CoV) and SARS-CoV-2, and ORF8b for Middle East respiratory syndrome CoV (MERS-CoV). Previous studies have shown ORF8b and ORF9b have immunoevasive properties, while the only known information for MHV I is its localization within the virion of the hepatotropic/neurotropic A59 strain of MHV. Whether MHV I is an innate immune antagonist or has other functions has not been evaluated. In this report, we show that the I protein of the neurotropic JHM strain of MHV (JHMV) lacks a N terminal domain present in other MHV strains, has immunoevasive properties, and is a component of the virion. Genetic deletion of JHMV I (rJHMV^IΔ57-137) resulted in a highly attenuated virus bothin vitroandin vivothat displayed a post RNA replication/transcription defect that ultimately resulted in fewer infectious virions packaged compared with wild-type virus. This phenotype was only seen for rJHMV^IΔ57-137, suggesting the structural changes predicted for A59 I altered its function, as genetic deletion of A59 I did not change viral replication or pathogenicity. Together, these data show that JHMV I both acts as a mild innate immune antagonist and aids in viral assembly and infectious virus production, and suggest that the internal N proteins from different betacoronaviruses have both common and virus strain-specific properties.IMPORTANCECoV accessory genes are largely studied in overexpression assays and have been identified as innate immune antagonists. However, functions identified after overexpression are often not confirmed in the infected animal host. Furthermore, some accessory proteins are components of the CoV virion, but their role in viral replication and release remains unclear. Here, we utilized reverse genetics to abrogate expression of a conserved CoV accessory gene, the internal N (“I”) gene, of the neurotropic JHMV strain of MHV and found that loss of the I gene resulted in a post replication defect that reduced virion assembly and ultimately infectious virus production, while also increasing some inflammatory molecule expression. Thus, the JHMV I protein has roles in virion assembly that were previously underappreciated and in immunoevasion. 

Freshwater scarcity is a global problem that requires collective efforts across all industry sectors. Nevertheless, a lack of access to operational water footprint data bars many applications from exploring optimization opportunities hidden within the temporal and spatial variations. To break this barrier into research in water sustainability, we build a dataset for operation direct water usage in the cooling systems and indirect water embedded in electricity generation. Our dataset consists of the hourly water efficiency of major U.S. cities and states from 2019 to 2023. We also offer cooling system models that capture the impact of weather on water efficiency. We present a preliminary analysis of our dataset and discuss three potential applications that can benefit from it. Our dataset is publicly available at Open Science Framework (OSF). 

This paper studies online resource allocation with replenishable budgets, where budgets can be replenished on top of the initial budget and an agent sequentially chooses online allocation decisions without violating the available budget constraint at each round. We propose a novel online algorithm, called OACP (Opportunistic Allocation with Conservative Pricing), that conservatively adjusts dual variables while opportunistically utilizing available resources. OACP achieves a bounded asymptotic competitive ratio in adversarial settings as the number of decision rounds T gets large. Importantly, the asymptotic competitive ratio of OACP is optimal in the absence of additional assumptions on budget replenishment. To further improve the competitive ratio, we make a mild assumption that there is budget replenishment every T* ≥ 1 decision rounds and propose OACP+ to dynamically adjust the total budget assignment for online allocation. Next, we move beyond the worst-case and propose LA-OACP (Learning-Augmented OACP/OACP+), a novel learning-augmented algorithm for online allocation with replenishable budgets. We prove that LA-OACP can improve the average utility compared to OACP/OACP+ when the ML predictor is properly trained, while still offering worst-case utility guarantees when the ML predictions are arbitrarily wrong. Finally, we run simulation studies of sustainable AI inference powered by renewables, validating our analysis and demonstrating the empirical benefits of LA-OACP. 

We study online convex optimization with switching costs, a practically important but also extremely challenging problem due to the lack of complete offline information. By tapping into the power of machine learning (ML) based optimizers, ML-augmented online algorithms (also referred to as expert calibration in this paper) have been emerging as state of the art, with provable worst-case performance guarantees. Nonetheless, by using the standard practice of training an ML model as a standalone optimizer and plugging it into an ML-augmented algorithm, the average cost performance can be highly unsatisfactory. In order to address the "how to learn" challenge, we propose EC-L2O (expert-calibrated learning to optimize), which trains an ML-based optimizer by explicitly taking into account the downstream expert calibrator. To accomplish this, we propose a new differentiable expert calibrator that generalizes regularized online balanced descent and offers a provably better competitive ratio than pure ML predictions when the prediction error is large. For training, our loss function is a weighted sum of two different losses --- one minimizing the average ML prediction error for better robustness, and the other one minimizing the post-calibration average cost. We also provide theoretical analysis for EC-L2O, highlighting that expert calibration can be even beneficial for the average cost performance and that the high-percentile tail ratio of the cost achieved by EC-L2O to that of the offline optimal oracle (i.e., tail cost ratio) can be bounded. Finally, we test EC-L2O by running simulations for sustainable datacenter demand response. Our results demonstrate that EC-L2O can empirically achieve a lower average cost as well as a lower competitive ratio than the existing baseline algorithms. 

The condensation of baryons within a dark matter (DM) halo during galaxy formation should result in some contraction of the halo as the combined system settles into equilibrium. We quantify this effect on the cuspy primordial halos predicted by DM-only simulations for the baryon distributions observed in the galaxies of the SPARC database. We find that the DM halos of high surface brightness galaxies (with Σ eff  ≳ 100  L ⊙ pc −2 at 3.6 μm) experience strong contraction. Halos become more cuspy as a result of compression: the inner DM density slope increases with the baryonic surface mass density. We iteratively fit rotation curves to find the balance between initial halo parameters (constrained by abundance matching), compression, and stellar mass-to-light ratio. The resulting fits often require lower stellar masses than expected for stellar populations, particularly in galaxies with bulges: stellar mass must be reduced to make room for the DM it compresses. This trade off between dark and luminous mass is reminiscent of the cusp-core problem in dwarf galaxies, but occurs in more massive systems: the present-epoch DM halos cannot follow from cuspy primordial halos unless (1) the stellar mass-to-light ratios are systematically smaller than expected from standard stellar population synthesis models, and/or (2) there is a net outward mass redistribution from the initial cusp, even in massive galaxies widely considered to be immune from such effects.