[Paper Review] Generative AI and the Digital Commons

Saffron Huang, Divya Siddarth|arXiv (Cornell University)|Mar 20, 2023

Scientific Computing and Data ManagementDecision Sciences21 citations

TL;DR

This paper analyzes how Generative Foundation Models (GFMs) rely on the digital commons and proposes governance, data-sharing, and funding approaches to protect and sustain the commons while mitigating risks. It emphasizes consortia, data contribution norms, and input-data governance to align GFM development with public benefits.

ABSTRACT

Many generative foundation models (or GFMs) are trained on publicly available data and use public infrastructure, but 1) may degrade the "digital commons" that they depend on, and 2) do not have processes in place to return value captured to data producers and stakeholders. Existing conceptions of data rights and protection (focusing largely on individually-owned data and associated privacy concerns) and copyright or licensing-based models offer some instructive priors, but are ill-suited for the issues that may arise from models trained on commons-based data. We outline the risks posed by GFMs and why they are relevant to the digital commons, and propose numerous governance-based solutions that include investments in standardized dataset/model disclosure and other kinds of transparency when it comes to generative models' training and capabilities, consortia-based funding for monitoring/standards/auditing organizations, requirements or norms for GFM companies to contribute high quality data to the commons, and structures for shared ownership based on individual or community provision of fine-tuning data.

Motivation & Objective

• Assess how GFMs interact with and impact the digital commons
• Identify risks GFMs pose to the digital commons and democracy
• Propose governance structures, including consortia and data-contribution norms, to sustain the commons
• Suggest models for funding and shared ownership that reward data producers and contributors
• Outline practical pathways to monitor, audit, and standardize GFM practices

Proposed method

• Define GFMs and map their reliance on digital commons infrastructure and data sources
• Analyze risks to the digital commons including information quality, democracy, and labor dynamics
• Propose governance mechanisms such as consortia, data-contribution norms, and input-data governance
• Evaluate pros/cons of policy options like data dividends, copyright/licensing, and individual data rights
• Recommend implementation strategies for monitoring, auditing, and standards-setting bodies

Experimental results

Research questions

• RQ1 How do GFMs depend on and potentially undermine the digital commons?
• RQ2 What governance models can mitigate risks to the commons while promoting responsible GFM development?
• RQ3 What are the comparative advantages and drawbacks of consortia, norm-based data contributions, and input-data governance?
• RQ4 How can data producers be rewarded or compensated within a commons-focused framework?
• RQ5 What structures are feasible for monitoring and auditing GFMs in practice?

Key findings

• GFMs both depend on and may erode the digital commons through rapid generation and deployment of AI outputs.
• Open questions about compensation for data producers and the commons highlight governance gaps and potential monetization tensions.
• Consortia for monitoring, auditing, and standards-setting can coordinate multi-stakeholder input to manage risks.
• Norms for GFM companies to contribute high-quality data to the commons can improve data integrity and governance.
• Governance structures based on input data for model training can align incentives among data providers, researchers, and users.
• The paper discusses multiple policy options (data dividends/taxes, stricter copyright, individual data rights) with varying feasibility and trade-offs, emphasizing the need for infrastructure and governance to monitor effects.