The Label Blog

The landscape of artificial intelligence (AI) development is rapidly evolving, presenting both unprecedented opportunities and challenges for society. As AI becomes increasingly integrated into various aspects of our lives, there’s a growing recognition that its development cannot be left solely to commercial interests. This realization has given rise to the concept of Public AI, which aims to create an ecosystem of initiatives that promote public goods, public orientation, and public use throughout every step of AI development and deployment.

A multi-stakeholder approach is crucial for advancing Public AI, as no single entity can address all the complex issues surrounding AI development and deployment. This blog post will explore how different sectors of society can contribute to the advancement of Public AI, drawing insights from Mozilla’s recent report titled “Public AI: Making AI work for everyone, by everyone” (Mozilla, 2024).

Every sector of society—from developers and policymakers to academics, civil society organizations, companies, and the general public—has a crucial role to play in advancing public AI. By working together, we can create an AI ecosystem that serves the public interest and benefits everyone.

Developers: Building the Foundation

Developers are at the forefront of creating the technological foundation for Public AI. Their role is essential in building the models, tools, and infrastructure to power the next generation of AI applications. According to the Mozilla report, developers can contribute to Public AI in several key ways:

Creating open-source AI infrastructure

Developers should prioritize contributing to and maintaining open-source AI projects. This includes releasing pre-training datasets, pre-trained models, and tools under open licenses. By doing so, they help create a robust ecosystem of public goods that can serve as the backbone for a broader range of Public AI initiatives. The report highlights examples such as the Allen Institute for AI (AI2) and EleutherAI, which are dedicated to developing open-source AI models and tools (Mozilla, 2024).

Developing competitive alternatives to private AI

Developers must focus on building open and trustworthy alternatives to popular AI models, tools, and products. Just as Firefox provided an open-source alternative to Internet Explorer, Public AI needs competitive options that can pressure private competitors to adopt best practices while providing users with secure and private choices.

Integrating diverse perspectives in AI development

The Mozilla report emphasizes the importance of developers collaborating with experts, including legal scholars, social scientists, user experience designers, and trust and safety specialists. This interdisciplinary approach ensures that AI development considers various perspectives and potential societal impacts. Additionally, when building user-facing applications, developers should integrate best practices for co-design, including community-led need identification and participatory research.

Policymakers: Shaping the Ecosystem

Policymakers play a crucial role in creating the regulatory and funding environment that allows Public AI to flourish. The Mozilla report outlines several key areas where policymakers can make a significant impact:

Funding open and shared AI infrastructure

Governments should invest in making shared computing, data, hosting, models, and tools available to nonprofits, academics, startups, and the public. This can involve direct government investment in building infrastructure (e.g., through national laboratories) or providing grants and research funding. The report cites Germany’s Sovereign Tech Fund as a model for funding open digital infrastructure (Mozilla, 2024).

Creating a level playing field through regulation

Policymakers should pursue strong antitrust, consumer protection, and other regulatory policies across all layers of the AI stack. This helps ensure that Public AI initiatives have a fair chance to emerge alongside larger commercial actors. The report also emphasizes the importance of increasing interoperability between different AI offerings, making it easier for Public AI alternatives to be used alongside private AI products and services.

Promoting high-quality jobs in AI

The Mozilla report recommends that policymakers implement policies to advance workers’ rights, bargaining power, and workforce development for all workers involved in AI development and deployment. This includes mandating fair wages and benefits for government-funded AI jobs, creating outreach and education programs for underrepresented groups, and establishing strong legal protections for unions organizing in the AI sector.

Academics: Advancing Research and Education

Universities and academic institutions have a vital role in advancing Public AI through research, education, and training the next generation of AI practitioners. The Mozilla report highlights several key areas where academics can contribute:

Focusing on neglected areas of AI research

Academics should prioritize research in areas that corporate labs aren’t incentivized to pursue, such as AI for low-resource languages and AI fairness, safety, and interpretability. The report emphasizes the importance of releasing open-source research artifacts (e.g., code, models, datasets) and publishing findings in open-access journals to ensure widespread accessibility (Mozilla, 2024).

Supporting public interest organizations

Building channels to Public AI careers

Universities should establish partnerships that support AI workers in joining public interest organizations. The report cites the example of how the U.S. Department of Veterans Affairs partners with medical schools to bring health professionals into the VA system. Similar programs could help funnel AI talent into public sector and nonprofit roles, countering the current trend of AI graduates predominantly joining private industry.

Civil Society Organizations: Ensuring Public Benefit

Civil society organizations play a critical role in ensuring that Public AI truly serves the public interest. The Mozilla report outlines several key ways these organizations can contribute:

Guiding AI applications for public use

Civil society organizations should share their domain expertise with Public AI developers and practitioners. This includes identifying community needs, advocating for ethical data collection, anticipating potential harms, and determining when AI is or isn’t the appropriate solution to a particular problem (Mozilla, 2024).

Contributing philanthropic funding

Philanthropies significantly impact the funding of Public AI efforts throughout the development and deployment process. The report suggests that philanthropies can leverage their relative speed and flexibility to direct funding toward earlier-stage Public AI ventures, “moonshot” approaches, and regions and domains that have not yet received public sector funding.

Highlighting the benefits and risks of AI

Civil society organizations should use their platforms and storytelling capabilities to highlight both the benefits of Public AI and the potential harms of overreliance on private AI. The report also emphasizes the importance of these organizations calling attention to risks and harms from Public AI where they exist, helping ensure that Public AI lives up to its highest ideals.

Companies: Collaborating for Public Good

While Public AI aims to provide alternatives to purely commercial AI development, private companies still have an important function in advancing the Public AI ecosystem. The Mozilla report outlines several ways companies can contribute:

Open-sourcing models and tools

Companies should release pre-trained models, libraries, tools, and datasets in compliance with the Open Source AI Definition maintained by the Open Source Initiative. The report emphasizes that this should be done without restrictive licenses that some major commercial players have embraced (Mozilla, 2024).

Supporting security and accountability work

The report recommends that companies fund bug bounties for developers who identify and fix vulnerabilities in their code, making the open-source ecosystem more secure and higher quality. Companies should also commit to safe harbor policies for good faith security research and share the results of their model capabilities and security evaluations to promote transparency across the ecosystem.

Sponsoring cross-sector collaborations

Companies can sponsor research collaborations between their in-house researchers and academics, leveraging each sector’s strengths and resources. The report suggests that this can range from one-off projects to major initiatives, such as contributing funding and computing resources to AI research centers at local universities.

The Public: Driving Force Behind Public AI

Ultimately, the success of Public AI depends on engagement and support from the general public. The Mozilla report outlines several key ways individuals can contribute to advancing Public AI:

Identifying priority use cases

The public should advocate to civil society and AI organizations for the AI applications that would help them most, identifying gaps and unfulfilled promises in current AI development. This consumer advocacy helps ensure that Public AI serves everyone, not just the needs of large corporations (Mozilla, 2024).

Holding AI organizations accountable

Individuals should call out the misuse and harm of AI, especially about their own communities. The report suggests this can take various forms, from partnering with journalists and engaging in grassroots activism to testifying to policymakers and demanding AI protections in collective bargaining agreements.

Contributing data, feedback, and support

The public can help improve Public AI products by sharing and annotating data for crowdsourced citizen science platforms, participating in user research studies, and contributing to democratic AI efforts to define shared norms and values. The report emphasizes that participation from underrepresented communities is important to creating Public AI that truly represents the public.

Synergies and Collaborations

The advancement of Public AI requires collaboration across all sectors. The Mozilla report emphasizes the importance of an international and multi-stakeholder approach to catalyze collaborations, share expertise and build alignment toward shared approaches, values, and missions.

The report provides several examples of successful collaborations, such as the BLOOM AI model released by BigScience, an international collaborative initiative led by HuggingFace. This project brought together researchers, developers, and organizations worldwide to create an open, multilingual AI model (Mozilla, 2024).

Another example is the Trust and Safety Tooling Consortium, a research effort aiming to map and steer the ecosystem of open-source trust and safety tools. This collaborative approach makes it easier and more accessible for smaller organizations to implement best practices in their technology and platforms.

Conclusion

Advancing Public AI is a complex, multi-faceted challenge that requires the active participation of all sectors of society. From developers building the technological foundation to policymakers shaping the regulatory landscape, from academics advancing research to civil society organizations ensuring public benefit, from companies collaborating for the public good to individuals driving demand and accountability – each group has a crucial role to play.

As we continue to grapple with the rapid advancement of AI technology, it’s clear that a purely market-driven approach is insufficient to ensure that AI benefits all of humanity. By embracing a multi-stakeholder approach to Public AI, we can work towards an AI ecosystem that is more open, accountable, and aligned with the public interest.

The Mozilla report serves as a call to action for all of us to engage in our respective roles. Whether you’re a developer, policymaker, academic, activist, business leader, or concerned citizen, there are concrete steps you can take to contribute to the advancement of Public AI. By working together, we can shape an AI future that truly works for everyone, by everyone.

References: [1] Mozilla, “Public AI: Making AI work for everyone, by everyone”, https://assets.mofoprod.net/network/documents/Public_AI_Mozilla.pdf, September 2024.