X Close

UCL Department of Science, Technology, Engineering and Public Policy

Home

Applied in Focus. Global in Reach

Menu

Archive for the 'Policy Impact Unit' Category

Unleashing the economic potential of UK manufacturing

By a.tacu, on 2 May 2024

Image of speaker presentingManufacturing has a pivotal role to play in building a thriving future UK economy which is resilient and can meet many of the increasingly pressing challenges facing UK society.  

But is this fundamental role fully understood by those outside the world of manufacturing? Attending ‘The Future of UK Manufacturing’ event earlier this month has prompted me to reflect on this question.  

Common cross-sectoral challenges and potential solutions 

Recent policy developments such as the UK’s Advanced Manufacturing Plan supported by £4.5 billion of funding for strategic manufacturing sectors, the Net Zero Strategy and the UK Net Zero Research and Innovation Framework point to the increasing awareness of the importance of manufacturing for the UK economy in recent years.  

While this is welcome, a number of challenges continue to hamper the ability of manufacturing to realise its true economic potential. 

One of the key apparent challenges is that, despite the fact that the UK is a global leader in innovation and research, this does not fully translate into economic value through industrial activity [1]. For example, in terms of number of research publications, the UK surpasses the US in per capita terms, but it lags behind in translating scientific knowledge into commercial success. A telling example is that the word ‘manufacturing’ is only mentioned once in the ‘Science and Technology Framework’, which risks creating the perception that the onus is on industry when it comes to scaling up new technologies. 

Although a constant stream of fundamental research is a crucial part of the innovation ecosystem, the ability to scale up lab-based demonstrations needs to become an equally prominent part of how research is undertaken in the UK to set the right conditions for success. Lessons can be learned from the example of the Oxford-AstraZeneca Covid-19 vaccine where the parallel research in immunology and in manufacturing, supported by the Vax-Hub, contributed to the speedy Covid-19 vaccine rollout.  

Another significant challenge for the UK manufacturing community is skills gaps, which amounts to between £7.7 and £8.3 billion in lost annual economic output [2]. Part of the difficulty in attracting talent to manufacturing careers is that manufacturing jobs are still associated with a traditional view of manufacturing roles as being manual and poorly paid. A study led by InterAct suggests that these perceptions can be changed by focusing on levers which have the potential to attract people to manufacturing careers such as flexibility and being part of the solution to many of the health-related, environmental and economic issues we are facing. 

Which leads us to one of the recurring themes that emerged from the discussions held during the event – the importance of storytelling and narrative setting. There was agreement that the UK manufacturing community should challenge outdated perceptions and create a positive narrative about the role of manufacturing that cuts across sectors and is clearly communicated to policymakers and those outside the world of manufacturing. Coalescing around a common strong narrative can support with ensuring manufacturing remains high on the policymakers’ agendas and can attract the skilled people it needs.  

So, what could this narrative be? One of the workshop sessions explored this exact question. A strong narrative should show how manufacturing can be at the forefront of creating good quality jobs and be a fundamental part of the UK’s future economic prosperity and national security.  

Early-stage R&D, which underpins manufacturing innovation, requires continued long-term funding support as businesses are often risk-averse and not incentivised to invest sufficiently at that stage. The Engineering & Physical Sciences Research Council (EPSRC) has been responding to this need through its manufacturing for the future research funding programmes, which is very welcome. I am left convinced that prioritising this type of investment is more important than ever and that, over the long term, these investments will more than pay for themselves in value returned to the UK. 

Context 

The EPSRC together with the High Value Manufacturing Catapult and the Institute for Manufacturing at the University of Cambridge organised ‘The Future of UK Manufacturing’ event in Sheffield. The event brought together academics, policymakers, innovation agencies and industry to review the current UK manufacturing landscape and to look ahead to future research and innovation priorities and opportunities. I attended this event as Policy Adviser for Vax-Hub Sustainable, one of the manufacturing research hubs funded by EPSRC and co-led by UCL Biochemical Engineering and the University of Oxford.   

Author’s note 

Written by Anca Tacu, Policy Impact Unit. With thanks to Jen Reed, Head of Policy Impact Unit, for her valuable contributions.  

References 

[1] Cambridge Industrial Innovation Policy. 2024. UK Innovation Report 2024. Available at: https://www.ciip.group.cam.ac.uk/innovation/the-uk-innovation-report-2024/  

[2] Policy Connect. 2023. Upskilling Industry: Manufacturing productivity and growth in England. Available at: https://www.policyconnect.org.uk/research/upskilling-industry-manufacturing-productivity-and-growth-england  

The importance of collaboration to advance digital health

By luis.lacerda, on 27 March 2024

Earlier this month the Government announced a £3billion+ package to update fragmented and outdated IT systems across the NHS and transform the use of data to ease administrative burdens. That same week, the Policy Impact Unit (PIU) hosted a roundtable on digital health in the UK, bringing together colleagues from across UCL (see co-authors) as well as visiting researchers from the FioCruz Oswaldo Cruz Foundation in Brazil.

FioCruz is a federal public research foundation working with academic autonomy under the Ministry of Health of Brazil which was responsible for coordinating the COVID-19 vaccination campaign. The Brazilian delegation were keen to hear about UK experiences on health digitisation, challenges and opportunities, as well as developing a deeper understanding of the context and evaluation of several commitments agreed under the Brazil-UK High-Level Strategic Dialogues from 2020, some of which focussed on health cooperation and were funded by the Official Development Assistance (ODA) [1].

The main challenges discussed in the meeting, in relation to the digitalisation of the NHS, were systems’ interoperability, training and workforce capacity. Although there has been a push towards the adoption of federated data platforms (FDP), which will sit across NHS trusts and integrated care systems allowing them to connect data they already hold in a secure and safe environment, these are still disjointed and connecting them relies on platform providers talking to each other, which often does not happen.

Common challenges: interoperability, training and workforce capacity

The adoption of new digital health approaches is also reliant on having trained healthcare professionals to understand the power of data and new technologies. Particularly in primary care and GPs it is essential to have digitally literate colleagues that can engage communities, be clear and transparent about how health data is used and input it correctly to build FDPs that can be further used for research and to invest on the health of the nation[2]. Programmes like the NHS “Developing healthcare workers’ confidence in artificial intelligence” and inclusive digital healthcare are important, because there is a risk that ambitions to digitise the NHS, which are well intended, could exacerbate existing health inequalities and exclude some groups.

Incidentally, there is still a lack of progress to de-identify General Practice data and address low levels of confidence in new technologies among diverse communities – such as highlighted in the Health and Social Care Committee’s recent evaluation. Trust can be undermined as is the societal buy-in needed to deliver on ambitions for a more digital NHS.

Opportunities and way forward: innovation in regulatory mechanisms

On the flipside, there is an opportunity to bring people in early on to discussions on how AI tools are being used in medical devices, and how to properly manage the balance of risk and benefits such technologies may bring. The recent launch of the UK Regulatory Science and Innovation Networks was discussed, as well as the launch of a MHRA roadmap to create a framework for medical devices in the UK. Patients, researchers and industry representatives being included in this process, and being clear about how data can be used for the purposes of research, poses a great opportunity to bring real impact to clinical practice in terms of diagnosis, treatment, and monitoring of diseases.

Including other global partners in this conversation is essential given the importance of sharing learnings in different contexts, but also given the increasingly important role of international recognition in the medical domain as a factor to evidence impact. Specifically for global issues such as AI and post-market surveillance, where it is very difficult for regulators to know how new tools will perform before they are deployed, there is now a chance to have new standards emerge to shape digital health strategies across countries. We hope that visits like this inspire colleagues to work collaboratively and look forward to hearing from FioCruz how their visit is supporting Brazilian policy decisions on the development of digital health strategies.

Authors Note

Written by Dr. Luís Lacerda, Policy Impact Unit and co-authored by Professor Amitava Banerjee, UCL Institute of Health Informatics, Professor Derek Hill, UCL Dept of Medical Physics & Biomedical Engineering and Professor Patty Kostkova, UCL Institute for Risk & Disaster Reduction.

References

[1] For a list of projects funded under the scheme, please visit https://devtracker.fcdo.gov.uk/

[2] A particular good example was the COVID-19 registry where data such as vaccination rates, long-covid reports were put together in the same place and from different countries.

Adversarial Attacks, Robustness and Generalization in Deep Reinforcement Learning

By Ezgi Korkmaz, on 20 December 2023

Reinforcement learning has achieved substantial progress on successfully completing tasks, from solving complex games to large language models (i.e. GPT-4) including many different fields from medical applications to self-driving vehicles and finance, by learning from raw high-dimensional data with the utilization of deep neural networks as function approximators.

The vulnerabilities of deep reinforcement learning policies against adversarial attacks have been demonstrated in prior studies [1,2,3,4]. However, a recent study takes these vulnerabilities one step further and introduces natural attacks (i.e. natural changes to the environment given that these changes are imperceptible) while providing a contradistinction between adversarial attacks and natural attacks. The instances of such changes include, but are not limited to creating a blur, introduction of compression artifacts, or perspective projection of the state observations at a level that humans cannot perceive the change.

Intriguingly, the results reported demonstrate that these natural attacks are at least equally, and often more imperceptible compared to adversarial attacks, while causing larger drop in policy performance. While these results carry significant concerns regarding artificial intelligence safety [5,6,7], they further raise questions on the model’s security. Note that the prior studies on adversarial attacks on deep reinforcement learning rely on the strong adversary assumption, in which the adversary has access to the policy’s perception system, training details of the policy (e.g. algorithm, neural network architecture, training dataset), and the ability to alter observations in real time with simultaneous modifications to the observation system of the policy with computationally demanding adversarial formulations. Thus, the fact that natural attacks described in [8] are black-box adversarial attacks, i.e. the adversary does not have access to the training details of the policy and the policy’s perception system to compute the adversarial perturbations, raises further questions on machine learning safety and responsible artificial intelligence.

Furthermore, the second part of the paper investigates the robustness of adversarially trained deep reinforcement learning policies (i.e. robust reinforcement learning) under natural attacks, and demonstrates that vanilla trained deep reinforcement learning policies are more robust than adversarially trained policies. While these results reveal further security concerns regarding the robust reinforcement learning algorithms, they further demonstrate that adversarially trained deep reinforcement learning policies cannot generalize at the same level as straightforward vanilla trained deep reinforcement learning algorithms.

This study overall, while providing a contradistinction between adversarial attacks and natural black-box attacks, further reveals the connection between generalization in reinforcement learning and the adversarial perspective.

Author’s Note: This blog post is based on the paper ‘Adversarial Robust Deep Reinforcement Learning Requires Redefining Robustness’ published in AAAI 2023.
References:
[1] Adversarial Attacks on Neural Network Policies, ICLR 2017.
[2] Investigating Vulnerabilities of Deep Neural Policies. Conference on Uncertainty in Artificial Intelligence (UAI).
[3] Deep Reinforcement Learning Policies Learn Shared Adversarial Features Across MDPs. AAAI Conference on Artificial Intelligence, 2022. [Paper Link]
[4] Detecting Adversarial Directions in Deep Reinforcement Learning to Make Robust Decisions. International Conference on Machine Learning, ICML 2023. [Paper Link]
[5] New York Times. Global Leaders Warn A.I. Could Cause ‘Catastrophic’ Harm, November 2023.
[6] The Washington Post. 17 fatalities, 736 crashes: The shocking toll of Tesla’s Autopilot, June 2023.
[7] The Guardian. UK, US, EU and China sign declaration of AI’s ‘catastrophic’ danger, November 2023.
[8] Adversarial Robust Deep Reinforcement Learning Requires Redefining Robustness, AAAI Conference on Artificial Intelligence, 2023. [Paper Link]