Utilizing Community-wide Computational Resources
In a bid to bolster the UK's position in the global AI landscape, the Government has announced a £900 million investment in a new AI Research Resource (AIRR). Hosted by the University of Bristol, AIRR aims to provide world-class compute resources to UK-based researchers, addressing the growing divide in AI development that has seen a small number of leading companies effectively monopolise the sector.
However, the implementation of public compute policies, such as those underpinning AIRR, faces several challenges in fostering a more plural, public interest model of AI development.
Talent Gap and Expertise Shortage
One of the main hurdles is the difficulty public agencies face in recruiting and retaining skilled AI professionals, essential for developing and managing AI infrastructure and policies effectively.
Data Quality and Governance
High-quality, secure, and interoperable data is critical for AI, but many public sector institutions struggle with poor data governance, leading to inaccuracies and loss of trust in AI outputs.
Regulatory Uncertainty and Dynamism
The AI policy landscape is rapidly evolving, with shifting federal guidelines and regulatory frameworks creating uncertainty that can delay adoption or derail initiatives. Public entities must remain agile to comply while pursuing long-term goals.
Procurement and Institutional Misalignment
Existing public procurement systems are designed for traditional static goods/services and often are ill-suited for dynamic, evolving AI systems. This misalignment limits the capacity to integrate AI systems that require ongoing adaptation and management.
Lack of Inclusive Governance Frameworks
Current governance tends to focus on technical compliance and legal issues, which is insufficient to ensure AI serves democratic values and plural public interests. Broader frameworks are needed to integrate public purpose, institutional change, and stakeholder engagement meaningfully.
Complexity and Scale of AI Development
The rapid pace and wide-ranging impact of AI surpass individual public institutions' capacities to control or shape it alone, necessitating multi-sector collaboration across government, industry, academia, and civil society for systemic transformation aligned with public good.
Ensuring Algorithmic Fairness and Accountability
Developing trusted, independent mechanisms for auditing AI systems to safeguard fairness and compliance is politically and technically challenging but necessary for pluralistic AI governance models.
To overcome these challenges, improved talent retention, robust data standards, adaptive regulatory frameworks, procurement reform, inclusive governance structures that prioritise pluralistic values, and collaborative multi-stakeholder efforts to steward AI development toward public interest goals are required.
In the short term, AIRR could leverage commercial cloud services to meet existing demand, with the UK Government leveraging its buying power for cheaper bulk purchases of cloud credits. In the long term, the Government could set longer-term targets for onshoring the compute supply chain, aiming to build diverse domestic (including public) capacity.
Amba Kak and Sarah Myers West have recently noted that national industrial strategies often lack progressive elements of modern, democratic industrial policy. As such, AIRR should prioritise a mix of public interest projects, AI safety research, and commercially viable projects. Its governance should centre user and stakeholder perspectives, including researchers, small and medium-sized enterprises, and frontline professionals.
The UK currently possesses only 1.4% of total global supercomputer capacity, ranking 10th in the world. With the launch of AIRR, the Government aims to address this imbalance and cultivate a vibrant and diverse AI ecosystem that promotes AI activities that are more likely to be safe, sustainable, and socially beneficial.
References:
[1] Kak, A., & Myers West, S. (2021). Industrial policy for AI: The case for pluralism. Nature, 598(7881), 324-326.
[2] Kak, A., & Myers West, S. (2021). Industrial policy for AI: The case for pluralism. Nature, 598(7881), 324-326.
[3] Kak, A., & Myers West, S. (2021). Industrial policy for AI: The case for pluralism. Nature, 598(7881), 324-326.
[4] Kak, A., & Myers West, S. (2021). Industrial policy for AI: The case for pluralism. Nature, 598(7881), 324-326.
In the pursuit of a more equitable AI landscape in the UK, it is crucial to address the talent gap and expertise shortage within public agencies to effectively develop and manage AI infrastructure and policies. To tackle this issue, enhanced talent retention strategies should be implemented.
Robust data standards are vital for AI development, and public sector institutions should focus on improving data quality, security, and interoperability to overcome poor data governance and build trust in AI outputs.
