Artificial Intelligence x Real Economy
By Daniel Wilson, Policy & Public Affairs Director, Innovation & International
This week, AI safety tops the policy agenda: ensuring the truly extraordinary possibilities heralded by a small number of frontier AI firms are realised without imperilling us all as citizens.
It’s a vital goal, and we support UK ambitions to lead on it.
But this important aim mustn’t eclipse the arguably bigger leadership opportunity: safely harnessing all manner of AI across all UK sectors and public services to improve everyday life.
The Real Opportunity
That’s because too often the AI debate focuses only on global developers and on protecting citizens, neglecting the intermediaries, or deployers of AI. Yet this ‘missing middle’ is where most of the UK’s businesses and public sector bodies sit, and where vast economic and societal value resides.
The potential is huge – but, based on the UK’s low rates of AI deployment, could go untapped.
While the Government rightly celebrates that UK AI startups raise more funding than their peers in Germany and France combined, deployment of AI by UK firms is lower than in either country, lagging Italy and Spain too. Meanwhile in public sector deployment, Singapore and the US race ahead.
In short, we have an AI deployment deficit – and it’s costing the UK, big time.
So what has stopped the UK from leading so far?
Barriers to deployment
Among the commonly identified checks on AI deployment, three stand out:
- Transparency and regulatory/legal certainty: firms can find it hard to understand what AI they’re buying, despite deployers carrying the most regulatory risk. Globally, around 3 in 5 companies identify ‘AI vendors who don’t include explainability features’ as a pervasive barrier to explainable and trustworthy AI. Regulatory/legal uncertainty can frustrate deployment too.
- Skills and confidence: many firms don’t have the skills among their leadership or employees to understand AI and its potential for them, or to deploy it with confidence. The proportion of firms reporting skills as a key barrier to AI adoption is almost 60% higher in the UK than in France.
- Data complexity and quality: companies also list data complexity among the principal barriers to AI deployment. While AI may now get all the limelight (UK Google searches for AI increased x5 in the first half of this year alone), it relies on good quality data. It doesn’t matter how mighty your model, there’s one unbending rule: “rubbish in, rubbish out.”
These factors –transparency/regulatory certainty, skills, and data complexity – all drag deployment.
Yet they can be overcome with the right strategic focus, regulatory tools, and partnership between government and industry.
We don’t have all the answers (you can read more about BT Group’s own aspirations to responsibly embrace AI here). But we do have some proposals that should help.
We recommend an immediate focus on the following areas.
1. Transparency to de-risk deployment: to give companies confidence to buy and deploy AI, developers should have to publish consistent, easy-to-use information. The format should be a standardised ‘model card’ that explains what training data was used, how the AI system was tested, if and how bias mitigation was applied, and the limits to its recommended use. No proprietary information should be disclosed. Indeed, model cards should remain as simple as possible, like food labelling that lists ingredients, allergens and simplified nutritional ratings. This approach would help ensure responsibilities are distributed fairly across the value chain and promote clearer accountabilities. Greater transparency is a better alternative at present to a detailed UK legal framework for such fast-evolving technology.
2. AI ready workforce: while Government has helped expand and diversify the pipeline of new AI talent, more focus is needed on making the existing workforce ready for AI in the workplace. For individuals, there should be more flexibility in when they access Government support for post-18 learning, opening more opportunities to upskill and reskill, including in roles that AI can augment or create. For corporates, there should be more flexibility in how they invest their apprenticeship levy, enabling them to scale up programmes for current staff. For smaller enterprises, managers should get support in learning about how AI can help their business. Here again, flexibility in support can be a key factor in adoption.
3. Get the data foundations right: unnecessary uncertainty hangs over UK data laws. We’ve seen a stop/start approach to sensible reforms aimed at maintaining a high standard in data protection while allowing businesses to simplify their processes. Now’s time to finish the job and look to where pro-innovation provisions offer the most potential for AI-driven value. This could start with better sharing of Government data to improve public service delivery. We also need to ensure that ‘smart data’ schemes for industry focus on innovation between sectors, not narrowly within them. For example, ‘Open Communications’ proposals – focused only on telecoms – would currently provide poor bang-for-buck (Government’s own figures suggest costs could be more than double the consumer value unlocked). These measures should be part of a wider, ongoing strategic focus on data as well as AI; the fuel as well as the machinery of our present industrial revolution.
These policy proposals should help ensure that UK firms and institutions have better understanding, a better skills base and stronger data foundations upon which to build their AI adoption plans. They should contribute towards the resilience of the UK AI economy in the face of supply chain and labour market disruption. Crucially, they should benefit UK businesses, workers and consumers.
The safety of Frontier AI is a necessary pillar of trustworthy AI, but it is not sufficient for the challenges we face in adopting these technologies, responsibly, at scale, for the benefit of all. As the last visitors exit Bletchley Park this week, we hope the UK can enter a new phase focused on responsibly harnessing AI for growth. Making Artificial Intelligence work for the real economy. Making the safe adoption of AI a national mission.