Data Governance in the Age of Large-Scale Data-Driven Language Technology

By Yacine Jernite, Huu Nguyen, Stella Biderman, A. Rogers, Maraim Masoud, V. Danchev, Samson Tan, Alexandra Sasha Luccioni, Nishant Subramani, Gérard Dupont and 11 more. 2022

65 citations recorded by the source corpus.

The recent emergence and adoption of Machine Learning technology, and specifically of Large Language Models, has drawn attention to the need for systematic and transparent management of language data. This work proposes an approach to global language data governance that attempts to organize data management amongst stakeholders, values, and rights. Our proposal is informed by prior work on distributed governance that accounts for human values and grounded by an international research collaboration that brings together researchers and practitioners from 60 countries. The framework we present is a multi-party international governance structure focused on language data, and incorporating technical and organizational tools needed to support its work.

Subfields: Artificial Intelligence, Safety Research, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Safety Research

In Safety, Robustness & Interpretability

Citations in the corpus

BLOOM: A 176B-Parameter Open-Access Multilingual Language Model (2022, background): “Progress on the first goal is presented in the work of Jernite et al. (2022), which further motivates the needs and requirements of data governance, and outlines the structure needed for a network 4. bigscience.huggingface.co/blog/bigscience-ethical-charter of data custodians, rights-holders, and…”
Evaluating the Social Impact of Generative AI Systems in Systems and Society (2023, background): “The immense size of training datasets [118] makes scrutiny increasingly difficult.”
Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling (2023, methodology): “, 2022) The ROOTS dataset that BLOOM was trained on is available via application to researchers, but the authors suggest that they may not make the full data indefinitely available in accompanying work (Jernite et al., 2022; McMillan-Major et al., 2022).”
Foundation Models and Fair Use (2023, background): “The recent literature on data governance and curation discusses fair use in machine learning (Jernite et al., 2022; Paullada et al., 2021; Ganguli et al., 2022).”
The Stack: 3 TB of permissively licensed source code (2022, background): “For example, one could argue that code creators should have the rights to meaningfully control whether their data is included in the training set (Jernite et al., 2022).”
The BigScience ROOTS Corpus: A 1.6TB Composite Multilingual Dataset (2023, methodology): “, 2022), and data governance issues (Jernite et al., 2022) with the explicit goal of doing good for and by the people whose data we collected.”
Auditing large language models: a three-layered approach (2023, background): “, interactive model cards [183] and information about the initial training dataset [184], [185], to help downstream developers adapt the model appropriately.”
The future landscape of large language models in medicine (2023, background): “For example, they can break down complex concepts to an amateur level (Box 2, Supplementary Data, example 8, 9) and provide individualized feedback on academic topics with reasonable explanations (Supplementary Data, example 9) 6 .”
Pile of Law: Learning Responsible Data Filtering from the Law and a 256GB Open-Source Legal Dataset (2022, background): “The presence of private and toxic content in the most popular corpora for pretraining large language models is a well-known problem [11, 58].”
BloombergGPT: A Large Language Model for Finance (2023, methodology): “This data may be cleaned or subsetted in various ways before use (Touvron et al., 2023; Rae et al., 2020; Scao et al., 2022; Jernite et al., 2022), but issues of data duplication (Carlini et al.”
Harms from Increasingly Agentic Algorithmic Systems (2023, background): “[97] highlight the imbalance of power between AI developers and the rest of society, and propose frameworks and”