Sectoral Data Bargaining

History suggests that following significant technological breakthroughs, the economic contributions of individuals and communities can become more interchangeable, forcing them to endure temporary but damaging losses of bargaining power.

Similarly, AI’s benefits will likely accrue eventually to whole societies, but destabilizing short and medium-term disruptions could significantly offset those benefits. It is therefore important to strategize toward achieving social equilibrium quickly and robustly, by means that support the processes of technological development.

Large-scale cooperation to produce and control unique, industry-leading datasets for AI training and fine-tuning is the best way to rebuild bargaining power and demonstrate the importance of human contributions to the AI economy. Collective organization of data collection and management can and should play a powerful role in AI’s future. If successful, the creation of unique datasets will enable unique capabilities, resulting in AI progress that is both faster and more responsible, just as it has in past technological revolutions. It will also give genuine leverage to such initiatives.

This requires multilateral and sectoral strategies that unite open source models and closed collectively-controlled data.

No single actor—whether governments, technologists, or civil society—can counterbalance AI’s disruptive concentration of power. But together, we can build a new framework for responsible and powerful AI development, guided by these principles:

Support open source AI models within the bounds of safety.
Establish “trusted data intermediaries” (TDIs)—entities that negotiate on behalf of individuals in the data marketplace or develop their own models using high-quality, collectively-produced, non-public data.
Enact the right for every individual to opt out from TDIs.
Encourage cultural, political, and economic initiatives that foster the production of meaningful datasets, managed by TDIs.

The RadicalxChange Foundation and its partners have been working to develop real-world applications to illustrate what successful initiatives pursuing this strategy could look like.

First, by pioneering a trusted data intermediary in the cultural sector: Through Serpentine Arts Technologies’ collaboration with artists Holly Herndon & Mat Dryhurst, and partnerships with the Centre for Data Futures, King’s College London, RadicalxChange, AWO and Keystone Law, we have created a legal and governance framework to steward a dataset of UK choir audio recordings for AI model training. Trusted Data Intermediary for UK Choral AI Dataset & Trusted Data Intermediaries - RadicalxChange

Second: RadicalxChange and its partner OpenMined Foundation are building a Trusted Data Intermediary to help news organizations (re)gain bargaining power, specifically when negotiating access to their proprietary data with LLM companies. We propose to make available a custom tool enabling journalistic organizations or their sectoral-level representatives to manage third-parties’ permissions to train and fine-tune AI systems on proprietary news data. Saving Journalism with Sectoral Data Bargaining: A Pilot & MyData Matters Blogspot: Saving Journalism

Sectoral Data Bargaining

Further Reading

Donate via: