AI in the UK Insights from the IoD and techUK roundtable

The following outlines the key points raised during the discussion which was split into two sections. Firstly the meeting discussed the opportunities for greater AI adoption by all UK businesses and the challenges that may be preventing this from happening before turning to the regulation of AI in the UK and how to get this right for AI developers and users.

AI Adoption by UK industry – Opportunities and challenges

Right from the outset it was acknowledged that 2023 has been a breakthrough year for AI in terms of technical capabilities, public awareness and AI regulatory proposals. It was even suggested by attendees that the advent of ChatGPT had been the iPhone moment for AI.

These AI tools are already starting to be used to drive efficiencies, but even greater benefits are just on the horizon. In the near future, we can expect new transformative business models and companies as a result of AI advancements. Attendees recognised the huge opportunities that AI could bring, both now and in the near future, but also recognised the importance of ensuring we get AI governance and regulation right and support British enterprises that are currently falling behind when it comes to AI adoption, application and use.

On the topic of challenges, techUK initially highlighted the five current barriers to AI adoption that were summarised in techUK’s AI Adoption in the UK report earlier this year, these were: inconsistent data quality and accessibility, lack of trust in AI, a limiting organisational culture and understanding, insufficient compute infrastructure and an AI skills gap.

Participants then unpacked some of the most critical challenges impacting AI adoption by UK industry. Key issues raised to be addressed in order to encourage greater AI adoption included the following:

• Need to recognise there is a “missing middle”. Participants acknowledged that there was a group of small to medium-size enterprises throughout the UK that are already falling behind when it comes to AI adoption. This group was identified as the “missing middle”. These are businesses that are currently struggling with basic enabling actions such as access to talent, data and compute, as well as issues such as a lack of available funding and compelling business cases. The group emphasised the need for government to focus on supporting this group through access to the technology in an affordable and resilient way, and the importance of highlighting the commercial benefits that come from AI adoption and use through case studies. Access to advice and support on how to adopt AI was also deemed key to increasing AI adoption.
• Need for clear definitions and a common, shared language. One participant suggested that there was a real appetite to focus on AI technology but that the current dialogue/terminology being used in the AI debate is very confused especially with vendors using different terminology. There is therefore a need for a common language and clear definitions to prevent people across the supply chain from talking at cross purposes.
• The UK’s AI commercialization challenge. Whilst the UK has a strong academic and research footing when it comes to AI, converting these discoveries into tangible commercial outcomes has often been a struggle for the UK, said one participant. It was suggested that further work is needed by government to address this challenge of helping to convert leadership in academic research into successful commercial products and services.
• Need for greater transparency. Working with standards bodies, sector regulators need to ensure developers adopt standards to provider greater detail in terms of how their model has been trained, how the LLM model was tested and, if and how bias mitigation has been applied, this would help promote greater transparency and provide trust and confidence across the AI supply chain that the model has been developed responsibly.
• Need for legal clarity to build market confidence- One participant highlighted that a significant barrier for their organisation was a lack of clarity from government in providing a legal route to market for their AI product. This lack of clarity on how the company can move forward provides uncertainty in terms of how they will commercialise. This may force some companies to go elsewhere, such as the US, where legal clarity may be more readily available.

AI regulation in the UK

During the second half of the session the roundtable shared views on how to make AI regulation work in practice for businesses and the UK Government’s approach outlined in the AI white paper. In particular attendees discussed whether existing sector regulators have the right resources to fulfil the new roles and responsibilities outlined in the white paper, and the extent to which further legislation may be necessary.

• UK vs. EU approach to regulating AI- Participants acknowledged that the UK’s risk-based, context-specific, agile and outcomes-focused approach to governing AI could be a competitive advantage if implemented successfully. It was felt that the risk with the EU’s approach is that it’s an AI-specific piece of legislation, focusing on regulating a set definition of AI, which could easily become outdated. However, whilst it’s still unclear how the EU’s approach will be implemented at a member state level, there is a risk that the EU’s approach may become the de facto gold standard approach given the pace it’s developing. The UK must continue to stay on course, and work at pace to ensure the ambitions of the UK AI Whitepaper are seen through to fruition.
• Need for clear government leadership on AI – One participant highlighted that there has been a lot of important groundwork that has been done across government including with the UK AI Whitepaper, AI Assurance roadmap, and the work of the CDEI and Office for AI. But there currently exists a “fuzziness” in terms of who within government leads and coordinates AI policy initiatives across government.
• Highlighting the importance of the new Central Function- Participants discussed the new Central Function proposed in the white paper. This was seen as key to supporting the different regulators governing AI, and reducing the risk of regulatory overlap or duplication. In addition for the UK’s approach to work in practice greater resourcing and capability at the regulatory level was seen as key, as well as a central repository of expertise within the Central Function.
• Strengthening the capabilities of the new Frontier Model Taskforce – One participant also raised the need for the new Frontier Model Taskforce to be armed with deeper technological capacity to help specifically address some of the biggest challenges associated with frontier models. There was also a call for greater clarity on how the initial £100m of funding will be allocated, and whether this will be enough to sufficiently address arising challenges in this area.

Finally, the group recognised that considered regulation can ultimately spur innovation activity and provide companies with the confidence they need to effectively implement the technology, especially for the “missing middle” who don’t necessarily have the resource to take on liability risk.

Next steps

Attendees agreed that there is more that the IoD and techUK could be doing together with its members on understanding and realizing the opportunities and benefits of AI to UK businesses across all sectors and industries. This could include the sharing and promotion of real life use cases of AI in action and meetings where industry and companies can come together to discuss possible solutions to addressing the challenges facing firms in AI adoption.

techUK and the IoD committed to discuss in more detail what follow up joint activities and events could be in the coming weeks and months.