Artificial Intelligence Opportunities Action Plan

Lord Markham Excerpts

Thursday 16th January 2025 Lords Chamber

[Lord Markham (Con)]

I welcome the Secretary of State’s Statement in this space, and I start with an apology. When I agreed to speak to this, I was told it would be first business after Questions, and I am afraid I have to leave for a flight midway through, so I apologise to noble Lords and hope that they understand. My colleague will be here all the way through.

As I say, we welcome the Statement and we welcome the Matt Clifford plan, which my noble friend Lord Camrose kicked off when he was leading these efforts in government, so we see this as a positive step forward. As Health Minister during that time, I saw first-hand the potential of AI, how it can really transform our services and how the UK really does have the potential for a leadership role.

Matt Clifford's plan, we believe, is right that the role of the Government in this is really to establish the foundations for growth: namely, making sure we have an AI-skilled workforce, the computing power, the energy needs to drive that computing power and the right regulatory framework. Then we use the assets we have, such as the data, to create the right datasets and use our public sector to help the rollout in many of them. I will focus my comments and questions on how we are going to make sure that those things happen.

Turning to the first one, the AI-skilled workforce, I must admit that when I read in the report that 5,000 AI jobs were being created in this, like most of us, I thought “5,000—that is great.” Then you realise that, actually, 4,500 of those are in construction and only 500 are in AI itself, and you start to get worried that maybe this is a bit style over substance. I am very keen to understand from the Minister here what we are specifically doing in this space. I am mindful, for instance, that we talk about having government develop training for the universities with a delivery or reporting date of autumn 2027. We all know how quickly AI is moving in this space, and we are saying we are just going to have the training in place for the universities to give these courses in two and a half years’ time. I think we all know that, in two and a half years’ time, the world will have moved on massively from that, and no doubt the training will be out of place. I hope the Minister can come back on that and give us some reassurances that we will actually have an accelerated process—I am afraid this will be a bit of a recurring theme.

On computing power, my noble friend Lord Camrose, when he was in government, had secured an £800 million commitment to build a supercomputer in Culham. Now I read, in the Government’s action plan, that they will

“start to develop the business case process”

for an AI computer. Unfortunately, like many noble Lords, I know what that means: a Treasury business case process, so you are talking about a year and a half to two years, at least. All I can guarantee is that, if you take that length of time to produce a business plan, whatever you were planning in terms of a supercomputer will be superseded by advancements and events. What is the Minister doing to streamline that business plan process and get action on this front so that we can get that new supercomputer fast?

On energy, we all accept there is a desperate need for energy; again, that is laid down in the action plan. The Government’s answer to that is to set up an AI energy quango. I think most of us would say that we need to set out what our energy needs require, but then surely it is up to the network or GB Energy to fulfil that. Why do we need another quango and another layer of bureaucracy? What powers is that quango going to have if it will not be commissioning these facilities, which I assume GB Energy will do?

On regulation and governance, the regulatory framework is another very important part of the foundation. I know the Government have plans for an AI Bill, but what is the timeline for it? Again—this is a recurrent theme—it needs to be quick so we can keep up with events.

Moving on to AI datasets, I know that this is something that the Minister is very keen on in the health space, as am I, being the former Health Minister responsible for this area. We have the best health data in the world; the beauty of having a National Health Service is that we have data on primary and secondary care going back to the Second World War. We have data coming in from the UK Biobank and other sources, such as retina scans from opticians which, we are hearing, can be used for stroke detection or maybe the early warning signs of dementia. There are fantastic opportunities for this, and we can already see its applications around the health service today. We have been doing the research with focus groups to bring the public with us on the use of their healthcare data. We have the potential to create the UK Silicon Valley in the life sciences on the back of the data that we have. We had in place a data for R&D programme, which was looking to utilise and create datasets in the health space. Could the Minister update us on where we are with that, and whether it is going to be his focus? As we discussed, that is something I would be very happy to work on together.

The last part of the foundations is to use the assets that we have in the public sector as a rollout plan for that and, again, health is a perfect place for this. We have seen brilliant applications already in cancer treatment and in overprescriptions; there are possibilities with the NHS app, which is really taking off, and to use AI in the 111 service to help triage; these are all fantastic opportunities. We put in place an NHS productivity plan which was very AI driven and AI heavy. Could the Minister update us on the AI productivity plan for the NHS and what progress we are making on it?

To conclude, we are very positive about the opportunities AI provides to transform the whole country’s economy and public services in ways that we cannot even imagine. However, it is businesses that need to drive this. It is the role of the Government to set the foundations to allow business to deliver; it is not the role of quangos, which are not going to deliver it. This area will need a Minister to drive it through and make it happen. Is the Minister the one who will do that? If he is, I give him all our support and wish him the best of luck with it.

[Lord Clement-Jones (LD)]

My Lords, I also welcome this plan, perhaps with rather less baggage than the Conservative Benches. The Prime Minister and the Secretary of State invoked Babbage, Lovelace, Turing, the pioneering age of steam and even the white heat of the technological revolution, but at its core there is an important set of proposals with great potential. However, it is a wish list rather than a plan at present.

I particularly welcome the language in the plan around regulation, particularly where it refers to regulation assisting innovation, which is a change of tone. However, the plan and Statement raise many questions. In particular, how will the Government ensure that AI development mitigates risks beyond just safety to ensure responsible AI development and adoption, especially given the fact that a great deal of UK development will involve open-source applications?

On the question of the introduction of AI into the public sector, the Government are enormously enthusiastic. But, given their public sector digital transformation agenda, why are the Government watering down citizens’ rights in automated decision-making in the Data (Use and Access) Bill?

We welcome the recognition of the need to get the economic benefits for the UK from public sector data which may be used to develop AI models. What can the Minister tell us at this stage about what the national data library will look like? It is not clear that the Government yet know whether it will involve primary or secondary legislation or whatever. The plan and response also talk about “sovereign compute”, but what about sovereign cloud capability? The police cannot even find a supplier that guarantees its records will be stored in the UK.

While the focus on UK training is welcome, we must go beyond high-level skills. Not only are the tech companies calling out for technical skills, but AI is also shaping workplaces, services and lives. Will the Digital Inclusion Action Committee, chaired by the noble Baroness, Lady Armstrong, have a role in advising on this? Do the changes to funding and delivery expected for skills boot camps contribute to all of this?

On the question of energy requirements for the new data centres, will the new AI energy council be tasked with ensuring that they will have their own renewable energy sources? How will their location be decided, alongside that of the new AI growth centres?

The plan cannot be game-changing without public investment. It is about delivery, too, especially by the new sovereign data office; it cannot all be done with private sector investment. Where is the public money coming from, and over what timescale? An investment plan for compute is apparently to be married to the spending review; how does a 10-year timescale fit with this? I am very pleased that a clear role is identified for the Alan Turing Institute, but it is not yet clear what level of financial support it will get, alongside university research, exacompute capacity, and the British Business Bank in the spin-out/start-up pipeline support. What will the funding for the Compound Semiconductor Applications Catapult and the design and manufacturing ecosystem consist of?

The major negative in the plan for many of us, as the Minister already knows, is the failure to understand that our creative industries need to be able to derive benefits from their material used for training large language models. The plan ominously recommended reforming,

“the UK text and data mining regime so that it is at least as competitive as the EU”,

and the Government have stacked the cards in the consultation over this. We on these Benches and the creative industries will be fighting tooth and nail any new text and data mining exemption requiring opt-out.