Written Evidence Submitted by National Physical Laboratory

(GAI0053)

Context to NPL’s response

1. The National Physical Laboratory (NPL)[1] is the UK’s National Metrology Institute (NMI), responsible for developing and maintaining the UK’s primary measurement standards. NPL is a Public Sector Research Establishment (PSRE), which works in partnership with government, academia, applied research laboratories, public sector, and industry. NPL is part of the National Measurement System (NMS) which provides the UK with a national measurement infrastructure and delivers the UK Measurement Strategy on behalf of the Department for Business, Energy and Industrial Strategy (BEIS). 

2. AI and Machine Learning (ML) will have applications across a broad range of sectors, from healthcare to autonomous vehicles to climate monitoring. NPL supports the development of trustworthy AI/ML and provides confidence in the use of data through its deep expertise in metrology (the science of measurement)

3. NPL is a member of the AI Standards Hub[2], led by the Alan Turing Institute. The AI Standards Hub focuses on the role that standards can play in trustworthy and responsible deployment of AI and is a key component of delivering the UK’s National AI Strategy[3].

4. NPL is a member of the Alliance for Data Science Professionals[4],[5]and together with our partners - the Institute of Mathematics and its Applications, the Royal Statistical Society, the Operational Research Society, the BCS, the Alan Turing Institute, the Royal Society and the Royal Academy of Engineering we are contributing to the development of industry standards and accreditation for the data science profession.

How effective is current governance of AI in the UK?

5. There must be agreement on the components of good governance of AI that will cover at a minimum- safety, resilience, transparency, accountability and ethical considerations. The AI Standards Hub is working with the wider community to lead the conversations required to reach a consensus on specific sector and application requirements, particularly across different risk profiles, for example for safety critical decisions in healthcare.

6. There are significant benefits to bringing together experts from a wide range of areas to help establish a pro-innovation approach to regulating AI and to inform and advise on the governance of AI. Measurement expertise alongside that of colleagues from the UK’s wider National Quality Infrastructure including the British Standards Institution and UK Accreditation Service can add value, cutting across sectors and AI applications, from healthcare to autonomy.

7. Certification of AI products will be one of the key mechanisms in AI assurance, along with having the correct standards in place and a clear strategy for how AI will be regulated in the UK. Measurement expertise, infrastructure and standards are critical to underpin regulation and certification.

What measures could make the use of AI more transparent and explainable to the public?

8. NPL is undertaking research to enable trustworthy applications of AI. Measurement is key to building the foundations for the assurance of AI, understanding the processes and decisions. Internationally, NPL is leading work to develop quality frameworks, metrics, traceability chains and quantification of uncertainty in machine learning artefacts.

9. Public trust and confidence in AI and its use will rely on transparency in a number of aspects of its operation, including: how it is applied, when it is applied, why it is applied, the data on which it relies on to make decisions, and how that data is collected, processed and how the results are interpreted.

10. For AI to be deployed in appropriate situations successfully in a safe and ethical manner, the wider landscape and systems that it interacts within, must be taken into account.

Standards

11. The AI Standards Hub is led by the Alan Turing Institute in partnership with the British Standards Institution (BSI), National Physical Laboratory (NPL) and the Office for Artificial Intelligence – a joint unit between the Department for Digital, Culture, Media and Sport (DCMS) and the Department for Business, Energy and Industrial Strategy (BEIS).

12. The AI Standards Hub provides a platform to ensure that industry, regulators, academic researchers and the public have the tools and knowledge they need to contribute to the development of standards and make informed use of published standards to advance trustworthy and responsible AI. The Hub has a repository of information that includes in addition to the standards that are published, draft standards that are under development, providing the opportunity for input and comment.

13. Measurement standards play a key role in demonstrating reliability and trustworthiness by ensuring traceability back to national standards and by quantifying the uncertainty associated with the data to give a measure of its reliability.

14. A key challenge is that the demand for AI is at different stages of maturity in different sectors. Standards will play an increasing role in the governance of AI as more and more applications of AI reach the market.

Standards for data scientists, analysts and professionals

15. NPL is a member of the Alliance for Data Science Professionals. The Alliance is developing standards and providing professional recognition for colleagues working in a variety of data science roles.

16. The development and provision of professional standards is important for public confidence in the people responsible for the ethical, safe and resilient use of data, ensuring that standards are met, that data is being kept securely, analysed robustly and used in an ethical manner.

(November 2022)