Invest in Public AI and Building a Coalition of Open Democratic Economies
Two Pieces on How Carney Should Orient A Newly Hopeful, United and Ambitious Canada
Returning to Canada after 3 months of travel, it feels like the country has turned a historic corner. Unified by a trade war, there is hope in a new Liberal government which has rallied Conservative leaders like Ontario’s Premier.
This new government should:
- focus innovation and digital economy investments on Public AI, and the open source-, sovereign- and public digital- infrastructure and data ecosystems necessary to support it.
- rapidly grow Canada’s multilateral relationships, particularly in trade, research, foreign investment and aid, with like-minded values-aligned countries such as Japan, Scandinavia, Taiwan, Germany, Australia, New Zealand, the Netherlands, the UK and France, Spain and Italy ideally to create a coalition of these liberal democratic economies
Below I quote a manifesto by
advocating for the coalition and then include excerpts from a forthcoming paper by (SFU), (Mozilla) and me on Canada as a Champion for Public AI.Please reach out if you’d like to support either or both proposals.
Jon Shell’s Manifesto for the Rest(o)
Since I learned from my father about Canada’s multilateral leadership under Lester B Pearson in the creation of the International Declaration of Human Rights and under Lloyd Axworthy as the leading soft and middle power in securing treaties on land mines and CFCs, I’ve dreamt about returning our country and the world to similarly positive and ambitious missions. Jon Shell proposes exactly this in his brilliant manifesto:
“The short version: 10 countries, aligned and similar in many ways, have as much economic power and natural resources as either the US or China. Working together, they could form a new Alliance to counter the superpowers both militarily and economically, to protect democracy and to work toward social and environmental sustainability. They could lead the building of a new world, safe for our children and our grandchildren. Right now, it’s the only project that matters.
[…]
the fraying started before Donald Trump. Across the Western countries that formed the post-war coalition inequality has been on the rise and faith in capitalism and democracy in decline. Most no longer believe that children will be better off than their parents and young urban workers are furious that they can’t afford housing. Authoritarian political parties are feeding off the anger and hopelessness, often blaming those very same institutions for the challenges we face.
[…]
This article is an attempt to imagine a world where “middle powers,” like my home country of Canada, work together to check the influence of China and the US. Where countries with similar approaches to democracy, social welfare and public safety can protect their prosperity and security in a world otherwise governed by raw power. And where hope for a better future for our children can be restored.
[…]
As the world re-arms, America is the dominant supplier of military technology, creating obvious risks when Trump admits they may supply inferior equipment because “someday maybe they’re not our allies.” Much of the western world uses American technology platforms in their work and home lives, creating data risks and making it harder to build and keep home-grown technology companies. The vast majority of new funding for AI technology is flowing to the US. And the current administration seems focused on supercharging capital flight to America, with policies like the $5M “gold” visa, slashing taxes on corporations and the wealthy, and, of course, tariffs.
It is now impossible to imagine a better future with America playing as large a role as it does today. Extraordinary effort has been put into international agreements like the Paris Accord and the Global Minimum Tax to have them rendered toothless after being abandoned by the US. As long as we depend on American leadership for action, it will be hard to face other threats that require collective action, like climate change, inequality, and the increasing influence of the tech billionaires on elections around the world.
[...]
More hopefully, it is also suddenly more possible than ever to imagine a world where the US plays a less dominant role. Europe is gearing up to defend Ukraine — with or without the US. Canadian patriotism has reached a level that this Canadian did not think possible. Many Global South regional alliances are asserting greater influence and a desire for autonomy. Leaders across the world are expressing deep anger and betrayal at the actions of this administration. Some have declared the old world to be at an end. If there ever was a time to act and change our current course, it’s now.
[...]
Today, a set of wealthy countries with similar approaches to taxation, social welfare and public safety collectively rival the resources of both the United States and China. The five largest EU economies (Germany, France, Italy, Spain and the Netherlands) plus Japan, the UK, Canada, South Korea and Australia (we’ll call it the “Core 10”) represent about a quarter of the global economy, compared to about 26% for the US and 17% for China, and almost 600 million people. They all have relatively strong democracies⁴ and generally low levels of corruption.⁵ They are good places to live and do business. Collectively, they possess significant reserves of natural resources, a strong manufacturing base, several global centres of finance, some of the world’s best universities and access to critical shipping lines.
While these countries have all had strong economies for a while, it’s critical to understand how unusual this is. The world tends to be dominated by a few major powers: think the Soviet Union and the US, the colonial powers, Persia, Rome. There has rarely been as many rich and independent countries with similar approaches to governance as there are today, controlling this much territory and wealth. There’s almost a default to see the world through a “superpower” lens, with the current version as the US vs. China. The rest of us sit on the sidelines, exerting influence where possible, but knowing the outcome is not in our hands.
[...]
The purpose of the Alliance is to save and protect democracy and multilateralism for generations to come and to ensure that superpowers no longer control global affairs. Free from the dominance of the United States and China, the Alliance will imagine new possibilities, with the collective influence and capacity to convince others. The more independent and resilient this Alliance of like-minded countries is, the more stable and safe the world will be.
[...]
Three commitments that could help a new alliance to ensure an independent future are to collectively be responsible for ensuring the security and prosperity of its members against both military and economic aggression, to cooperate in efficiently and judiciously developing, transporting and refining natural resources and to work and invest collaboratively in technology, governance and scientific discovery to support the resilience of its economies, enhance its security and improve the lives of its citizens.
[...]
A series of creative and aggressive moves in the short term, even prior to signing any sort of treaty, would demonstrate that the middle powers were independent and intent on a coordinated response. A very public meeting could be held in the short term, ending with a signed declaration to work toward a new Alliance. Prospective members could act to coordinate their foreign aid to replace some of the work of USAID, immediately growing soft power and embarrassing the US around the world. Non-European prospective members could make firm and specific commitments to join Europe in the defense of Ukraine. Prospective members could announce a flashy partnership, like a space mission.
Other measures could demonstrate the repercussions of recent American policy. For example, a partnership to build a munitions factory to supply European countries in Quebec to take advantage of both efficient energy supply and a surplus of aluminum now subject to unilateral American tariffs. Or deals to switch suppliers of military equipment from the US to companies in prospective member states.
On “Liberation Day” Robert Reich recommended a series of actions for Canada, Mexico, Japan, the UK and the EU, including tariffs, preventing American banks from accessing their stock markets and increasing taxes and regulations on American digital platforms. Cory Doctorow has a much more fun suggestion to allow the “jail-breaking” of Teslas outside the US so its owners can access all its services for free. At SCP we crowd-sourced a pile of ideas. With creativity and coordination, the rest of us could create the same kind of chaos for America as they’re creating for us.
[...]
Whatever specific path it takes, the Alliance could kick off decades of innovation to rival the period following WWII. Our imagination is currently shackled by the interests of superpowers and monopolists and working within a global framework that no longer seems fit for purpose. Donald Trump has removed the constraints that governed American policy. If we remove our own constraints, the rest of us can take advantage of the opportunity to work toward military, economic and technological independence, and to imagine a different world.
Find the full post here: https://www.linkedin.com/pulse/manifesto-resto-jon-shell-wa05f
DRAFT:
Canada as a Champion for Public AI: Data, Compute and Open Source Infrastructure for Economic Growth and Inclusive Innovation
This paper makes the case for Canada investing in — and building more international collaboration around — public AI, if we want to meet our innovation and economic development goals.
It's a working draft
, and I recently presented at a conference on internet policy. Grateful if you would circulate widely. Please do get in touch if you agree with it or are working on related projects. We’re looking for allies and critiques and to build the coalition of people, countries, companies and organizations that can make a Public AI alternative a reality.ABSTRACT
Current AI models and products are predominantly built by a very small number of actors within only two incentive systems and political contexts, namely big tech, centred in Silicon Valley and China. This poses serious concerns about how AI will impact the rest of the world. The concentration of influence could prevent building AI as a public good, inhibit nations like Canada from unlocking transformative economic change, and increase risks, from democratic deficits, to greater inequality, to existential concerns, worldwide. In this article, we outline a tractable roadmap for Canada to work with a coalition of public and private actors around the world to champion a third option: “Public AI”. These actors should include governments institutions, private funds and companies, research institutes, open source projects, and civic organizations. This proposal builds on others’ extensive research on AI as a public good or utility, such as proposals for a CERN or Airbus for AI. If successful, this proposal would allow Canada, in conjunction with public and private allies around the world, to build and deploy competitive public AI infrastructure that drives economic growth and public benefit. Publicly backed AI research and compute programs, as well as commercial and nonprofit open source AI projects, already provide a solid foundation. What we need is: 1) greater and faster international collaboration, 2) explicit, publicly backed open source and standards-based strategies that make it possible for innovations and improvements in one country, company, or lab to accrue into a common pool of interoperable public goods used by all actors, and 3) explicit public data strategies that enable AI development and public interest uses of AI.
PAPER
[...]
The internet, the web and the open source software stack that currently underpins most digital businesses and public services [...] created widely spread social and economic benefits in many parts of the world.
[...]
the current AI landscape is dominated by the market – and by a very small number of companies – in a way that may short circuit the potential economic and social benefits, as seen with earlier eras of digital technology. This structure makes it less likely for AI development to prioritize safety, democracy, trust, healthcare and education relative to their societal value.
[...]
The Our vision for public AI in this article – inspired heavily by the Public AI Network [24] and Mozilla and Columbia University’s recent work on open source [9], responsible data [8] and AI safety [2] – is a robust and geographically distributed ecosystem of initiatives that create public goods, have a public orientation, and facilitate public use throughout every step of AI development and deployment, including compute, data, tooling and talent.
[...]
Canada is ideally positioned to help build a coalition of governments, companies, non-profits, and open source projects investing in and stewarding public AI, based on its history of AI innovation, international coalition building, and leadership in areas like human rights and multiculturalism
[...]
By pursuing “digital sovereignty through collaboration” – pooling resources with others to create an "Airbus or CERN for AI", ideas for which other scholars have developed substantive and tractable proposals (outlined below) – Canada can help establish a third pole in AI development distinct from the big tech and China, enable pluralistic uses of AI, and secure both economic independence and social benefit.
[...]
public AI means building AI systems that are made available as public goods (like utilities, roads, or open source software [53]), and these systems must maintain public accessibility, democratic accountability, and financial and environmental sustainability
[...]
The reality is that neither a single government in a country like Canada nor mid-sized tech companies can compete alone against the massive AI investments of big tech and China. The solution is what we might call digital sovereignty through collaboration: pooling resources with like-minded governments, companies, academia, civil society and open source projects to create collective capacity.
This is the Airbus model - named after the European aircraft manufacturer created as a multinational consortium to compete with American aviation dominance. An "Airbus for AI" would bring together democratic nations with shared values to build public AI infrastructure at sufficient scale to serve as a meaningful third pole.
[...]
Data represents a potential “Achilles Heel” [32] for the current generation of generative AI products. These models are trained on data acquired through potentially illegal, anticompetitive, and certainly unsustainable methods, from including scraping creative works without compensation to and harvesting personal information without meaningful consent [30]. In some jurisdictions, current commercial AI products could be deemed illegal because of their data dependencies or could be rejected by consumers on moral grounds. This backlash could elevate open-access and openly licensed LLM data-based public AI models to become the most competitive AI offering.
[...]
Another opportunity for public AI allies to advance public interest data would be building new data intermediaries [28]. Data intermediaries could help reshape the flow of data between users and companies. For instance, when a user sends a query to an LLM or a post to social media, the actual content of that message could first be stored on a server operated by an intermediary, which governance processes to answer questions like: What types of data will be sent to private companies and what will be contributed to a commons? How will compensation work? An ecosystem of intermediaries would make it possible to honour creator preferences and compensate data creation while accumulating high-quality data in a public pool. Practically, intermediaries could be instantiated as public benefit companies, nonprofits or even governments in places where the rule of law is strong (the “intermediary” concept could actually involve multiple specific implementations, including trusts, brokerages, unions, and collectives).
[...]
Canada's climate, abundant clean energy (and long-term planning around energy), and world-class institutes like Mila, the Vector Institute, and Amii [10] all provide unique strengths for contributing towards compute pooling. However, this doesn't mean building all the capacity within Canada. Rather, we would contribute strategically to a distributed network that provides resilience, diversity, and scale far beyond what any single participant could achieve alone, building on existing initiatives in Europe and conversations like those at the Paris AI Action Summit.
Our partners should not just be European nor just nations: they should include countries across Asia and the global majority, large cities and regions, as well as non-state actors including corporations like Cohere and Hugging Face, experts in decentralization like the Ethereum Foundation and Project Liberty, foundations like the Future of Life Institute and key civil society actors from standards organizations to open source projects. To use one example, Canada should form a closer relationship with the leading chip manufacturer Taiwan, as we have discussed with its Tech Ambassador and former Minister of Digital Affairs Audrey Tang (who has written extensively on Plurality and decentralized collaboration [44]).
[...]
For public AI to succeed, we need to build a fully open stack for AI that makes these components available to anyone. This assertion is based on experience from the first era of the internet, where open source came to serve as the core underpinning of most infrastructure and a huge driver of economic growth. Over the last 20 years, the Linux stack has become the standard infrastructure upon which all things digital are built, underpinning 90% [46] of publicly available cloud computing services and running 100% of the top 500 super computers [38]. The core open source stack including Linux represents $4 billion in supply side value (how much it cost to create) and $8 trillion in demand side value (the value it created for businesses and governments) [23]. The technical and economic value created by open source in the last era of the internet emerged by many businesses and governments — including competitors — pooling resources to create robust core infrastructure that everyone can build on. Public AI needs to take a similar approach.
[...]
The good news is that many parts of the AI ecosystem are already open source — and we have seen a move towards open source in base models over the last two years (Hugging Face’s AI Action plan [34] and Stanford’s AI Index reports [50] provide coverage of these trends). This new wave of base models spans a spectrum of openness [2, 9] from those with openly licensed software, datasets and model weights (e.g., OLMo from the nonprofit AI2, the Allen Institute for AI [35] and Salamandra from BSC, the Barcelona Supercomputing Center [20]) to those with open model weights, training and fine tuning software but no access to the pretraining data (e.g. various models from Mistral) to those that simply offer open weights (e.g., Llama from Meta, though Meta has signalled [54] that they may not release the weights for future Llama models.). While the most rapid progress has been in open weight models like Llama, fully open models are catching up.
There are major upsides to investing in a fully open AI stack and major risks from relying on proprietary players. Having a fully open stack – with all training code and documentation alongside data for pretraining, post training, and evaluation – provides the freedom to fork any element of the stack. For example, it becomes possible to build a model deeply tuned to a specific cultural or language context by intervening at a particular checkpoint. A fully open stack also provides the transparency to fully scrutinise a trained model for cultural or language bias, for hidden security threats, or simply to better understand model behaviour. It also makes it easier for many different parties to contribute back to a common stack as they fork and innovate in their own context.
[...]
5. CONCRETE ACTIONS FOR CANADIAN PUBLIC AI LEADERSHIP
To push forward the vision outlined above, Canada should: (1) assemble a public-private, international, and interprovincial public AI collaboration strategy, (2) champion an internationally focused open source AI strategy, and (3) build a public AI focused national data strategy.
5.1 International and Public Private Collaboration
Canada is well positioned to act as a catalytic and central force within the public AI coalition. Specific actions – each of which will be an undertaking – that Canada should take include:
- Convene an "Airbus for AI" consortium consisting of a core set of governments, companies and international and civil society organizations.
- Develop and champion a proposal to coordinate AI compute investments with the EU, other nations, key international players like the OECD, UN, IGF and non-governmental players like foundations
- Prioritize clean energy locations and the $15 billion the federal government proposed to match $30 billion in pension fund support to build Canadian data centres [51]. Make these available to consortium members at cost.
- Use public AI strategy to coordinate investments between Canadian provinces.
- Fund civic organizations, cultural institutions, and public services to work with their international and interprovincial peers on public AI initiatives.
[...]
5.2 Open Source Strategy
Canada should find ways to encourage researchers and companies to use and contribute to the broader open source AI stack:
- Open source artifacts funded by public dollars such as those produced by national labs, and public-private partnerships
- Work with allies around the world to develop practical ways to pool and rapidly improve the public AI stack through a joint commitment to open source, open content and open data
- Ensure that base models are kept truly open and unrestricted while also allowing elements of the stack to be used in more restrictive settings
- Encourage international collaboration between existing open source efforts. Open sourcing is already happening with contributions from Canadian firms like Cohere and Transformer Labs, European firms like Mistral, nonprofit labs like AI2 and Kyutai, and national projects like AI Singapore, AI Sweden, BSC, and the UK AI Safety Institute.
[...]
5.3 Data Strategy
Canada should establish a data strategy that at once enables rapid progress on AI innovation and at the same time prioritizes fair compensation, privacy and representation. Specifically, Canada should:
- Create and support both data trusts (looking to examples like the Department of Canadian Heritage’s framework [15]) and sectoral data cooperatives in areas of Canadian strength, such as healthcare, clean energy, and financial services.
- Create new publicly managed “data flywheels” that feed high-quality data into said trusts, cooperatives, and open source projects
- Fund institutions that are producing, archiving, and curating public domain and openly licensed data, looking to the examples described above like the Internet Archive.
- Implement frameworks for indigenous data sovereignty to ensure that indigenous communities maintain control over their own data.
- Lead in establishing cross-border data sharing agreements specifically for public AI development
6. CONCLUSION: THE TIME FOR CANADIAN LEADERSHIP IS NOW
The window for establishing alternatives to the current trajectory of AI development and to create economic opportunity for players beyond big tech and China is rapidly closing. As compute resources concentrate, data sources are exhausted or locked up, and technical talent flows to a handful of companies that are led by a small number of billionaires with decreasing democratic checks on their power, AI’s potential future benefits for the average human relative to its risks diminish.
Canada has the expertise, resources, and diplomatic credibility to champion an alternative approach, with benefits for Canada and for democratic societies worldwide. By focusing on public infrastructure, international collaboration, and balancing public and private interests, we can create a distinctive and economically viable position in the global AI landscape. For the Canadian economy, this will have immediate benefits: we can support Canadian-based AI companies, attract international talent, and reduce dependencies on foreign AI platforms and associated economic leakage. Further, we can embed Canadian values of democracy and respect
for rights into AI systems, preserve Canadian cultural sovereignty and dignity in an era of AI-driven content and services, and contribute broadly to making AI systems that lead to fair outcomes for diverse populations.
Acting as a champion for a public AI consortium is a pragmatic economic strategy for a mid-sized power navigating between technologically advanced, economic superpowers. By leveraging our unique advantages and addressing the fundamental challenges of the current AI paradigm, Canada can secure both economic benefits and social goods. By pooling resources across compute, software, and data layers, this consortium could achieve the scale necessary to present a meaningful alternative to existing models.
For Canada, this represents both a return to our strengths in multilateralism and an innovative approach to technological development. Rather than attempting to win a game designed for larger players, we can help change the game itself - establishing rules and infrastructure that better serve democratic values and diverse interests. Canada can demonstrate the viability of alternatives to data acquired through potentially illegal, anticompetitive, and unsustainable methods. It can better balance regulation with funding and flexibility for innovation and talent to compete.
Collaboration can overcome big tech and China’s concentration of power and money by connecting the many existing publicly backed and open source AI initiatives so that they add up to more than the sum of their parts. Values can be a greater draw for talent than dollars. Academia, government and civil society can work more effectively with companies to create value through carefully choreographed public-private partnerships and ecosystems than black and white zero sum competition with the private sector. While this pooling approach can help to mitigate many of the reasons an ambitious undertaking like building public AI might fail, there is still work to be done in terms of documenting specific obstacles that will arise and producing reactive plans (e.g., how can public AI proceed if current compute investments targets are not met?)
The time to act is now. With clear vision, strategic investment, and international leadership, Canada can help ensure that the future of AI serves not just market interests, but public good. In doing so, we can secure our place in the next technological era while staying true to our values of pluralism, inclusion, and shared prosperity. And we can convert one of the most influential AI communities into one of the most lucrative and socially beneficial AI economies in the world.