Artificial Intelligence (AI) has rapidly become a transformative force in our society, reshaping industries, economies, and daily life. However, the current AI landscape is predominantly driven by commercial interests, with a handful of tech giants leading the charge. This concentration of power and resources raises important questions about the future of AI development and its impact on society.
While LLMs have demonstrated impressive capabilities in natural language processing and generation, the intense focus on them may come at the expense of investing in and developing other potentially beneficial types of AI systems. It’s worth considering whether the resources and talent being poured into LLMs could yield a greater positive impact if directed towards AI applications in areas like healthcare, education, sustainability, or scientific research.
Additionally, the development of LLMs and AI in general has significant environmental implications that need to be taken into account. The compute-intensive training of large AI models consumes substantial amounts of energy, contributing to carbon emissions. As AI systems become more prevalent, we must assess and mitigate their environmental impact, striving to make the development and deployment of AI more sustainable. Ultimately, thoughtful consideration of the innovation trajectory and ecological footprint of AI will be key to ensuring it delivers broad benefits to society.
In a recent document titled “Public AI: Making AI work for everyone, by everyone,” Mozilla presents a compelling case for an alternative approach: Public AI. This concept envisions a robust ecosystem of initiatives that promote public goods, public orientation, and public use throughout every step of AI development and deployment. As we’ll explore in this blog post, the need for alternatives to commercial AI development is becoming increasingly clear, and Public AI offers a promising path forward.
The Limitations of Commercial AI
While commercial AI development has undoubtedly led to significant innovations, it comes with inherent limitations that can hinder the technology’s full potential to benefit society:
Profit-driven focus neglects important applications
The Mozilla document highlights a crucial issue: “an AI ecosystem powered exclusively by the market will prioritize a narrow set of profitable applications” (Public AI, p. 4). This profit-driven focus means that many critical AI applications, such as detecting illegal mining operations, facilitating deliberative democracy, combatting disinformation, or delivering medicine to remote areas, are often under-resourced relative to their societal value.
Concentration of power and resources
The current AI landscape is dominated by a small number of large tech companies with the resources to invest billions in AI research and development. This concentration of power raises concerns about the diversity of approaches and applications being pursued. As the document states, “We can’t rely on a few companies to build everything our society needs from AI, and we can’t afford the risk that they won’t” (Public AI, p. 4).
Lack of diversity in development and applications
Commercial AI development tends to reflect the priorities and perspectives of a relatively narrow group of developers and investors. This can lead to AI systems that fail to address the needs of diverse communities or, worse, perpetuate existing biases and inequalities. The document emphasizes the importance of “expanding who has the power and opportunity to build AI in the first place” (Public AI, p. 4).
The Promise of Public AI
Public AI offers an alternative vision for AI development that aims to address these limitations and create a more inclusive, equitable, and beneficial AI ecosystem:
Definition and Goals of Public AI
According to the Mozilla document, Public AI is defined as “a robust ecosystem of initiatives that promote public goods, public orientation, and public use throughout every step of AI development and deployment” (Public AI, p. 3). This approach seeks to create AI that is open, accessible, and focused on serving the needs of communities rather than solely pursuing profit.
Potential benefits for Society and the Economy
By promoting a more diverse and inclusive approach to AI development, Public AI has the potential to:
- Address critical societal challenges that may be overlooked by commercial interests
- Increase access to AI tools and technologies for a broader range of individuals and organizations
- Foster economic growth around the world without concentrating gains in the hands of a few large companies
Fostering innovation and competition
Public AI can create a more level playing field in AI development, allowing for greater innovation and healthy competition. By reducing barriers to entry and promoting open-source approaches, Public AI can encourage a wider range of actors to contribute to and benefit from AI advancements.
Examples of Public AI Initiatives
Several initiatives are already demonstrating the potential of Public AI:
BLOOM: An open, multilingual AI model
The document highlights BLOOM as a prime example of Public AI in action. Developed by BigScience, an international collaborative initiative led by HuggingFace, BLOOM is an AI model that supports 46 languages and 13 programming languages. Importantly, it is released under the Responsible AI License, making it open to research and commercial use under ethical AI principles (Public AI, p. 17).
Common Voice: Crowdsourcing multilingual speech data
Mozilla’s own Common Voice project is another excellent example of Public AI. This open-source platform crowdsources multilingual speech and text data, enabling more people to develop AI applications that represent the world’s linguistic diversity. Common Voice has been recognized as a Digital Public Good by the UN-backed Digital Public Goods Alliance (Public AI, p. 17).
Other case studies
- FASST: A program by the U.S. Department of Energy that leverages existing scientific data, expertise, and supercomputing resources to accelerate AI progress in critical fields like scientific discovery and national security.
- vTaiwan: An online citizen engagement platform in Taiwan that uses AI-powered tools to facilitate democratic input on public policy (Public AI, p. 17).
These examples demonstrate the diverse applications and approaches possible within the Public AI ecosystem.
The Three Pillars of Public AI
The Mozilla outlines three core goals that define Public AI initiatives:
Public goods: Creating open, accessible resources
Public AI aims to create “open, accessible public goods and shared resources at all levels of the AI stack” (Public AI, p. 13). This includes making infrastructure like computing, data, hosting, models, tools, and education available to nonprofits, academics, and startups. By doing so, Public AI can reduce barriers to entry and foster a more diverse ecosystem of AI development.
Public orientation: Centering community needs
Public AI should “center the needs of people and communities, especially those most underserved by market-led development” (Public AI, p. 13). This involves integrating mechanisms for public participation, responsiveness, safety, and accountability throughout the AI development and deployment process. By collaborating with nonprofits, universities, labor unions, and governments, Public AI initiatives can better understand and address the needs of diverse communities.
Public use: Prioritizing applications in the public interest
The third pillar of Public AI is to “prioritize enabling applications in the public interest, especially in important domains neglected by the private market” (Public AI, p. 14). This could include developing AI for health diagnostics, preserving indigenous languages, or addressing climate change – areas where commercial incentives may be insufficient to drive innovation.
Challenges and Considerations
While Public AI offers immense potential, it also faces several challenges that must be addressed:
Funding and sustainability
One of the primary challenges for Public AI initiatives is securing sustainable funding. As the document notes, “Public AI initiatives will require sustainable revenue models to exist in the long term and remain competitive with the private market” (Public AI, p. 18). This may involve a combination of government funding, philanthropic support, and innovative business models.
Ensuring responsible use and development
As with any AI development, Public AI must prioritize ethical considerations and responsible use. The document warns that “there is a risk of creating a successful open source AI commons without also advancing the mechanisms to ensure that it is used in a trustworthy manner” (Public AI, p. 18). Developing robust governance structures and accountability mechanisms will be crucial.
Balancing public and private sector roles
While Public AI aims to provide alternatives to commercial AI development, it’s important to recognize that both public and private sectors have important roles to play. The document emphasizes that “Public AI cannot be solely dependent on governments” and calls for a “resilient and pluralistic AI ecosystem” (Public AI, p. 18-19). Finding the right balance and fostering productive collaborations between public and private entities will be an ongoing challenge.
Conclusion
The case for Public AI is compelling. As AI continues to shape our world, we need alternatives to purely commercial development that can address critical societal needs, promote diversity and inclusion, and ensure that the benefits of AI are widely shared. Public AI offers a vision for a more open, accessible, and community-oriented approach to AI development.
By supporting and engaging with Public AI initiatives, we can help create an AI ecosystem that truly works for everyone. This may involve contributing to open-source projects, advocating for public funding of AI research, or simply choosing to use AI products and services that align with Public AI principles. As the Mozilla document concludes, “If AI is impacting every sector of society, everyone must also be involved in its governance and beneficial deployment” (Public AI, p. 27).
The future of AI is not predetermined. By embracing the concept of Public AI, we have the opportunity to shape that future in a way that reflects our shared values and aspirations. It’s time to put the public back in AI.
References: [1] Mozilla, “Public AI: Making AI work for everyone, by everyone”, https://assets.mofoprod.net/network/documents/Public_AI_Mozilla.pdf, September 2024.
