Michelle Lin

Irene Solaiman, Zeerak Talat, William Agnew et al. · allen-ai, cmu

Generative AI systems across modalities, ranging from text (including code), image, audio, and video, have broad social impacts, but there is no official standard for means of evaluating those impacts or for which impacts should be evaluated. In this paper, we present a guide that moves toward a standard approach in evaluating a base generative AI system for any modality in two overarching categories: what can be evaluated in a base system independent of context and what can be evaluated in a societal context. Importantly, this refers to base systems that have no predetermined application or deployment context, including a model itself, as well as system components, such as training data. Our framework for a base system defines seven categories of social impact: bias, stereotypes, and representational harms; cultural values and sensitive content; disparate performance; privacy and data protection; financial costs; environmental costs; and data and content moderation labor costs. Suggested methods for evaluation apply to listed generative modalities and analyses of the limitations of existing evaluations serve as a starting point for necessary investment in future evaluations. We offer five overarching categories for what can be evaluated in a broader societal context, each with its own subcategories: trustworthiness and autonomy; inequality, marginalization, and violence; concentration of authority; labor and creativity; and ecosystem and environment. Each subcategory includes recommendations for mitigating harm.

Roberto Verdecchia, Luís Cruz, June Sallou et al.

With the growing availability of large-scale datasets, and the popularization of affordable storage and computational capabilities, the energy consumed by AI is becoming a growing concern. To address this issue, in recent years, studies have focused on demonstrating how AI energy efficiency can be improved by tuning the model training strategy. Nevertheless, how modifications applied to datasets can impact the energy consumption of AI is still an open question. To fill this gap, in this exploratory study, we evaluate if data-centric approaches can be utilized to improve AI energy efficiency. To achieve our goal, we conduct an empirical experiment, executed by considering 6 different AI algorithms, a dataset comprising 5,574 data points, and two dataset modifications (number of data points and number of features). Our results show evidence that, by exclusively conducting modifications on datasets, energy consumption can be drastically reduced (up to 92.16%), often at the cost of a negligible or even absent accuracy decline. As additional introductory results, we demonstrate how, by exclusively changing the algorithm used, energy savings up to two orders of magnitude can be achieved. In conclusion, this exploratory investigation empirically demonstrates the importance of applying data-centric techniques to improve AI energy efficiency. Our results call for a research agenda that focuses on data-centric techniques, to further enable and democratize Green AI.

Xiaofeng Zhang, Michelle Lin, Simon Lacoste-Julien et al.

The widespread adoption of generative AI models has raised growing concerns about representational harm and potential discriminatory outcomes. Yet, despite growing literature on this topic, the mechanisms by which bias emerges - especially in unconditional generation - remain disentangled. We define the bias of an attribute as the difference between the probability of its presence in the observed distribution and its expected proportion in an ideal reference distribution. In our analysis, we train a set of unconditional image generative models and adopt a commonly used bias evaluation framework to study bias shift between training and generated distributions. Our experiments reveal that the detected attribute shifts are small. We find that the attribute shifts are sensitive to the attribute classifier used to label generated images in the evaluation framework, particularly when its decision boundaries fall in high-density regions. Our empirical analysis indicates that this classifier sensitivity is often observed in attributes values that lie on a spectrum, as opposed to exhibiting a binary nature. This highlights the need for more representative labeling practices, understanding the shortcomings through greater scrutiny of evaluation frameworks, and recognizing the socially complex nature of attributes when evaluating bias.

Pratinav Seth, Michelle Lin, Brefo Dwamena Yaw et al.

Millions of abandoned oil and gas wells are scattered across the world, leaching methane into the atmosphere and toxic compounds into the groundwater. Many of these locations are unknown, preventing the wells from being plugged and their polluting effects averted. Remote sensing is a relatively unexplored tool for pinpointing abandoned wells at scale. We introduce the first large-scale benchmark dataset for this problem, leveraging medium-resolution multi-spectral satellite imagery from Planet Labs. Our curated dataset comprises over 213,000 wells (abandoned, suspended, and active) from Alberta, a region with especially high well density, sourced from the Alberta Energy Regulator and verified by domain experts. We evaluate baseline algorithms for well detection and segmentation, showing the promise of computer vision approaches but also significant room for improvement.