Democratic AI to serve the public – OneProject.org

In a rapidly evolving landscape where artificial intelligence (AI) is increasingly influencing critical aspects of daily life—from job opportunities to loan approvals—there is growing concern about the lack of public oversight in the development and deployment of these technologies. The current AI ecosystem is dominated by a small number of corporations that prioritize shareholder interests over societal welfare. This situation has prompted calls for a more democratic approach to AI governance, one that empowers the public to shape the rules and objectives of AI systems that profoundly affect their lives.

The Case for Democratic Governance in AI

The argument for democratic governance in AI is rooted in the recognition that the technology has far-reaching implications for society. The CEOs of major tech companies are often caught in a competitive race, feeling pressured to innovate quickly to avoid losing market share. This urgency leads to a structural problem: the governance of AI is largely determined by corporate interests rather than the public good. The proposal is not merely for improved regulation but rather for a fundamental shift in how decisions regarding AI are made. The public should have a direct say in the governance of AI, ensuring that the technology aligns with collective societal values and needs.

One proposed solution is the establishment of a Global AI Assembly (GAIA), which would serve as an international body to facilitate democratic participation in AI governance. This assembly would operate on four key democratic claims, each reinforcing the others, and would aim to ensure that AI systems are aligned with the objectives of the public rather than just those of corporations or military entities.

The AI safety community has focused on the concept of 'alignment,' which refers to ensuring that AI systems pursue objectives deemed 'right' or 'safe.' However, the deeper question remains: whose objectives should these systems be aligned with? Currently, the primary stakeholders are enterprise clients, military organizations, advertisers, and wealthy consumers. The voices of the general public, who ultimately bear the consequences of AI deployment, are often marginalized.

Participatory Governance as a Solution

To address this imbalance, the concept of participatory governance is put forward. This approach involves structured deliberative processes where a representative group of citizens engages with evidence and trade-offs to make informed decisions. A prominent method within this framework is the citizens' assembly, which consists of a cross-section of the public selected by lottery. These assemblies have successfully tackled significant issues, such as marriage equality in Ireland and climate recommendations in France.

Citizens' assemblies can set the boundaries for AI development by defining unacceptable risks, establishing non-negotiable safeguards, and determining the transparency required from AI developers. The idea is that while technical experts write the code, the public should dictate the parameters within which that code operates. This would create a system where all AI companies adhere to binding rules, thus preventing a 'race to the bottom' where companies cut corners to gain competitive advantage.

The proposal also emphasizes the importance of changing the incentives that drive AI development. Currently, companies are incentivized to maximize profits, often at the expense of ethical considerations. If public interests were prioritized, AI could be directed towards beneficial outcomes, such as educational tools that genuinely promote learning or healthcare solutions that prioritize patient well-being over profit margins.

The Need for a Public AI Dividend

As AI continues to automate various sectors, there is a risk that economic benefits will concentrate in the hands of a few corporations and their shareholders. This scenario, described as 'techno-feudalism,' raises questions about the equitable distribution of wealth generated by AI technologies. A proposed solution is the implementation of a public AI dividend, which would allocate a share of AI-driven productivity gains to the public. This would allow communities to decide how to use these funds, whether for retraining displaced workers or investing in social infrastructure.

The concept of participatory budgeting has been successfully implemented in numerous cities worldwide, demonstrating that communities are capable of managing public funds effectively. A public AI dividend could empower communities to address pressing needs, thereby ensuring that the benefits of AI are shared more equitably.

The call for public equity in AI companies is another critical aspect of this democratic approach. Rather than merely regulating AI, the public should hold a governing share in companies that significantly impact society. This would ensure that AI development aligns with public interests, rather than being solely driven by profit motives.

Conclusion

The need for democratic governance in AI is urgent. As the technology continues to evolve and permeate various aspects of life, it is crucial that the public has a voice in shaping its trajectory. The establishment of a Global AI Assembly could provide a framework for participatory governance, ensuring that AI serves the collective good rather than the interests of a select few. By empowering citizens to set the rules and objectives for AI, society can harness the technology's potential while safeguarding against its risks. The public is ready for this change, and it is time to organize that sentiment into a political demand for a more equitable future in AI governance.