Clifford Chance

On 6 January 2021, the UK Government's AI Council (the Council) published its AI Roadmap (the Roadmap). The publication of the Roadmap swiftly follows the UK's Brexit transition period ending.

The Roadmap outlines 16 recommendations for the UK Government's strategic direction on AI. The Roadmap recommendations are targeted and suggest deeply ambitious and innovative policy and regulatory steps for the UK to further cement its position as a global AI leader.

The implementation of the Roadmap will be watched closely by the UK's EU friends following its exit from the bloc, particularly as the EU refines its approach to trustworthy AI in parallel.

The Roadmap is part of a robust package of AI and data related policy and strategic documents published in recent weeks, including the UK National Data Strategy and the UK House of Lord's recent paper, which warned that the UK has "no room for complacency" for AI.

Four pillars for AI success

The Council's Chair, Tabitha Goldstaub, said "we must increase investment where it is working and plan for the long term in a way that integrates ethics, security and social impacts." This trust and ethics-driven approach aligns in spirit with the proposed EU principles for trustworthy AI, where more concrete regulatory proposals are expected this year.

To keep pace with global competition, improve the lives of its citizens and bring economic benefits in the form of recovery and resilience, the Roadmap states that the Government must create a strong base of general support for AI, prioritise select markets and consider its options for collaboration in the international landscape.

The Council has suggested four pillars:

1.      Research, Development and Innovation.

2.      Skills and Diversity.

3.      Data, Infrastructure and Public Trust.

4.      National, Cross-Sector Adoption.

Research, Development and Innovation

The Roadmap states that AI's full potential will only be realised in the UK if the government continues to create conditions for industry leading research, development and innovation in AI for both the public and private sectors. As such, the Roadmap has three recommendations:

1.      Scale up and make sustainable public sector investment in AI: this will ensure consistent access to top talent from around the world and create new ways to find and bring researchers, disciplines and sectors together.

2.      Increase funding for The Alan Turing Institute: this will establish it as a truly national institute and provide it with the confidence to plan and invest in strategic leadership for the UK in AI research, development and innovation.

3.      Encourage 'moonshot' programmes and ensure they are challenge-led, high-risk and scalable: AI, by definition, has a high risk, high reward nature and this lends itself to goal-directed 'moonshot' programmes which require people to work across boundaries and existing organisational structures to build new relationships and networks in order to develop solutions to large challenges. Such 'moonshots' might include developing safe, ethical, explainable and reproducible AI which will accelerate its use across many sectors.

The Council's recommendation for targeted investment and funding for The Alan Turing Institute is particularly noteworthy, as it underlines the Roadmap's theme of developing and working with ethicists and technologists that have the requisite expertise to help ensure that the AI regulatory approach in the UK is appropriately calibrated.

Skills and Diversity

Mistrust in AI is often linked to a lack of education, insufficient skills, and a misunderstanding of what AI actually is.

2020 emphasised the ways in which AI can create or perpetuate human bias. However, it also underlined that deep-seated human bias exists separate from machines and that AI, if properly trained, has the potential to mitigate this bias.

The Roadmap's ultimate vision is for UK citizens to live confidently with AI and, for those who develop it or work with it, to have the best possible starting foundation.

In order to achieve this, the Roadmap has three recommendations:

1.      Make diversity and inclusion a priority: this will involve benchmarking and forensically tracking levels of diversity to make data-led decisions about where to invest and ensure that underrepresented groups are given equal opportunities.

2.      Commit to achieving AI and data literacy for everyone: this means educating the general public such that they understand the risks and rewards of AI and can be confident and informed users.

3.      Scale up and commit to an ongoing 10 year programme of high level AI skill building: this will result in new research fellowships, AI-relevant PhDs, and industry-led Masters and apprenticeships.

Following release of the Roadmap, Dame Wendy Hall of the University of Southampton said "it is vital we keep our foot firmly on the accelerator with regards to skills and diversity".

Data, Infrastructure and Public Trust

The Roadmap states that the UK should aspire to be the world leader in access to safe, quality data and must promote public scrutiny of and input into automated decision making to ensure trust in AI. To achieve this, the government will need to put in place both physical and social infrastructure which will include broadband networks, high performance computing capacity, governance and good practice guidelines. The aim is to reimagine data availability and governance to improve standard practice and public trust. The Roadmap has four recommendations to achieve this:

1.      Lead the development of data governance: the Government must continue to drive the good work done by regulators, so far, with respect to auditing AI for personal data protection, such that an actionable legal framework and guidelines for different data sharing structures, such as trusts, cooperatives and contracts, are put in place.

The Roadmap highlights that the UK should to develop a clear and flexible National AI Strategy that links general regulatory principles to industry-specific needs to ensure that governance is carried out in a context that will enable innovation and growth. Furthermore, additional work is needed to build a common language amongst data practitioners to build an understanding of the nature of datasets and how best to manage them.

Finally, the UK must continue to invest in data-specific regulatory capacity and algorithmic accountability, ensuring that funding is available to cover advisory, enforcement and regulatory landscapes.

2.      Ensure public trust through public scrutiny: the Roadmap acknowledges that there is a significant amount of public scepticism directed at AI and algorithmic applications and that, without public trust, widespread uptake of AI will be difficult to achieve. As such, it recommends encouraging public engagement and involvement, where possible, in the development and deployment of AI to ensure transparency and accountability. On a practical level, this will involve revealing the purposes and training data behind algorithms and public engagement with algorithmic impact assessments and ensuring that any enforcement body has the capacity and confidence to apply regulation to its full extent.

3.      Consolidate and accelerate the infrastructure needed to increase access to data for AI: this point is inextricably linked with the two above and involves investing in  physical, virtual and social infrastructure, supporting relevant organisations, pursuing initiatives for pump priming innovation and enabling safe data sharing between sectors and institutions for valuable uses.

4.      Position the UK with respect to other major AI nations: the Roadmap states that the UK should urgently and sustainably build upon its strong international position in terms of AI. This should be focused on strengthening existing partnerships with "like-minded" countries (the Council specifically indicates France and Canada as examples), and considering how to increase bilateral cooperation with countries such as the US and China.

National, Cross-Sector Adoption

The final pillar in the Roadmap proposes transitional and specific measures to boost initial take-up of AI in the crucial target sectors of business, start-ups, public sector, health and social care, climate change and defence:

1.      Increase buyer confidence in AI capability across all sectors and size of company: The Roadmap states that creating confidence in AI such that businesses actively seek out AI solutions is critical. Covid-19 caused the accelerated adoption of digital systems to enable working from home and the UK should seek to capitalise on this momentum by creating grants to access impartial advice on AI, encouraging further investment in AI by businesses and enabling businesses, and especially SMEs, to be smart consumers of AI systems.

2.      Support the UK's AI start-up vendor community: the UK must enable greater access to data, infrastructure, expertise, funds, talent and specialist knowledge for AI focused start-ups to help them develop into market leading AI companies.

3.      Enable robust public sector investments in AI: the Roadmap states that there is an urgent need to have a digitally and data enabled local and national government. This will be achieved by ensuring that public sector projects practice intelligent procurement of AI and build in the use of data, analytics and AI for the public benefit.

4.      Use AI to meet the challenges of net zero carbon emissions: the Roadmap highlights that technology is already an essential tool for innovating to reduce the environmental impact of human activity. As such, the UK should accelerate the use of AI for these purposes by building it into relevant 'moonshots' and incentivising AI companies to address net zero carbon emissions.

5.      Use AI to help keep the country safe: the UK should work with government departments and agencies to ensure that AI is available to assess and respond to modern defence and security threats and opportunities. This could be achieved by the creation of a new MOD agency for AI.

6.      Build on the work of NHSX and others to lead the way in using AI to improve outcomes and create value in healthcare: AI can help manage the complex information flow that is needed to integrate NHS systems and, as such, healthcare should be on the forefront of AI adoption. Indeed, in order to build a competitive advantage, the UK must create a clear data strategy for healthcare, develop new partnerships between AI driven SMEs and the NHS and educate healthcare professionals working with AI technologies.

Call to action

The UK government department responsible for digital, culture and media and sport matters confirmed that the Council's recommendations form the beginning of a wider conversation.

The AI Council, alongside the UK government's Office for AI, will arrange workshops in early 2021 to engage on these topics. This underlines the government's intention to carefully consider these recommendations and how they can translate into practice – a key priority for AI in 2021.

Stakeholders are invited to participate in these workshops and can register to do so here.