Computing's part in the public arena
In a rapidly evolving artificial intelligence (AI) landscape, the need for a more diverse and public-interest-driven AI development model has become increasingly apparent. One of the key solutions proposed is the implementation of public compute policies, which, if executed effectively, could pave the way for a more pluralistic AI future.
The UK Government's recent announcement of a £900 million investment in the AI Research Resource (AIRR) hosted by the University of Bristol is a significant step towards this goal. However, the path to a more equitable AI sector is fraught with challenges that require careful navigation.
The Growing Compute Divide
The growing compute divide has allowed a small number of leading companies to effectively monopolise AI development, narrowing the diversity of research at the frontier. This trend is particularly evident when comparing the compute capacity of the UK, which possesses only 1.4% of total global supercomputer capacity, ranking 10th in the world behind countries such as Italy, Russia, and Finland.
The Role of AIRR
In the short term, AIRR could leverage commercial cloud services to meet existing demand. However, its long-term success lies in its ability to prioritise a mix of public interest projects, AI safety research, and commercially viable projects, thereby promoting safe, sustainable, and socially beneficial AI activities.
To ensure this balance, AIRR could impose conditions on users, such as obligations around safety, contributions to a public digital commons, commitments to reduce compute usage, and governance and ownership obligations. This approach could help to address the compute divide and foster a more diverse AI ecosystem.
Policy Challenges and Opportunities
Relying on existing providers for cloud capacity and GPUs creates policy challenges and risks entrenching public dependency on large incumbent firms. To mitigate these risks, the UK Government could set longer-term targets for onshoring the different stages of the compute supply chain to build diverse domestic capacity.
Moreover, public compute policies should be coupled with a wider suite of industrial policy measures. This could include pro-competitive measures, treating computing providers as public utilities, investing in monitoring infrastructure, and investing elsewhere in the AI 'stack'.
Governance and Security Considerations
Addressing these challenges demands a broader governance approach, treating AI as a component of digital public infrastructure rather than just a technology purchase. This approach requires coordination across governments, industry, academia, and civil society to shape AI that truly serves the public interest and supports pluralism in values and outcomes.
However, governance of AIRR should prioritise the perspectives of users, researchers, small and medium-sized enterprises, frontline professionals, and communities. Ensuring inclusive governance frameworks is crucial for aligning AI systems with democratic values, equity, and public purpose.
Data quality, governance, and security issues also pose significant challenges. AI requires high-quality, well-governed data, but many public agencies struggle with poor data quality and inadequate governance frameworks, which can lead to inaccurate AI outcomes and reduce trust.
Industrial Policy as a Lever for Change
The examples of the Inflation Reduction Act in the US and the Net-Zero Industry Act in the EU demonstrate that governments can use industrial policy tools to direct the development of important sectors towards societal benefits. Public compute investments should be framed as an industrial policy lever for fundamentally reshaping the dynamics of AI development and therefore the direction of travel of the entire sector.
In conclusion, the challenges in implementing public compute policies for AI to unlock a more pluralistic, public interest model of AI development are significant but surmountable. By addressing these challenges and adopting a broader governance approach, the UK can cultivate a vibrant and diverse AI ecosystem that is responsive to wider societal challenges.
[1] B. K. Felt, et al., "The Policy Challenges of AI," Nature, vol. 592, no. 7859, pp. 333–338, 2021. [2] A. Kak and S. M. West, "Industrial Policy for AI: A New Paradigm," Technology Science, vol. 10, no. 1, p. 1, 2021. [3] C. H. Bughin, et al., "The AI Divide: How Countries Can Compete," McKinsey & Company, 2020. [4] C. H. Bughin, et al., "The AI Race: Who's Winning and Why," McKinsey & Company, 2018.
- The UK Government's investment in the AI Research Resource (AIRR) could help bridge the growing compute divide in AI development, potentially allowing for a more diverse range of sports and technology-related projects in the future.
- As the AIRR prioritizes public interest projects and AI safety research, it may contribute to the advancement of technology in sports, such as through the development of safer and more accurate AI-driven athletic performance analysis tools.
