The Government’s Vision Is Good But We Need Delivery on Science and Tech
We are entering a new era of science and technology. As the founder of world-leading AI company DeepMind, Demis Hassabis, said last week, the impact of general AI could ultimately rival the impact of electricity. The transformative nature of AI – as well as developments in biotech, climate tech, quantum and much more – means good technology policy will be critical to future prosperity. The urgency of this issue requires the increasing trends towards incrementalism in our politics to be reversed. In its place we need a radical new policy agenda that reshapes the state and makes harnessing the potential of technology its primary purpose.
In publishing a series of reports and reviews this week – including Sir Paul Nurse’s review of research, development and innovation, the Future of Compute Review, and the Science and Technology Framework policy paper – and announcing an extension of the Horizon Europe Guarantee scheme, the British government showed that it is edging towards this realisation. In comments accompanying the releases, Prime Minister Rishi Sunak stated his desire to make Britain a science superpower by 2030 as part of “our new national mission”.
This had echoes of a report we published a few weeks ago, A New National Purpose, co-launched by Tony Blair and William Hague. The essence of our message was that the UK needs to move fast or risk falling further behind the tech giants set to dominate the decades ahead: the US and China. On a domestic level, the translation is simple: make technology our new national purpose, or become poorer. The report set out a plan of action that was commensurate with the task ahead of us. It made the case for science and tech to become the central driving force of government, using tools such as sovereign general-purpose AI systems, increases to R&D investment and major reforms to planning, procurement and research to accelerate the speed at which this happens.
To this end, the vision that the prime minister sets out is welcome, just as his announcement of a new Department for Science, Innovation and Technology was a few weeks before. There were some good announcements: investment in AI, quantum and synthetic biology; new Focused Research Organisations; and more funding for AI PhDs.
But the reality is that government action needs to go much further to address systemic issues, as well as deeper in driving reform from the centre. We have seen vision from this government, but not enough in terms of delivery.
This is despite increasing warning signs, whether from the life sciences or concerns raised to a parliamentary committee about how hyper-cautious regulators are impeding the progress of our space industry. Tech companies are also being put off by the UK’s current lack of competitiveness. In the last week alone, chip designer Arm and building-materials company CRH have shunned the UK market, instead choosing the US for their IPOs. As we wrote in our report, this is an issue of growth equity, as well as around wider capital-market reforms.
A major issue in need of fixing is pensions. While the Science and Technology Framework’s broad commitment to engage with defined-contribution schemes to unlock investment into UK science and technology is welcome, there is a lack of detail about how ministers will seek to achieve this. As argued in our report, we need a dramatic consolidation of the UK pensions market; the current multitude of small schemes is making long-term, risk-friendly investment in science and technology firms more difficult. Tax changes to encourage the consolidation of these schemes would be the best way forward.
We also need to end Treasury micro-management of investment in science and technology. Investing in the key technologies of tomorrow is not the same as typical government investing, and requires a different approach that allows for the kind of long-term, riskier investments necessary to spur technological breakthroughs. Ministers should look to create a new science and tech policy and delivery unit across Number Ten and the Cabinet Office, and to exempt science projects from the kind of business-case, “value for money” criteria that act as brakes on high-risk, high-reward investment.
Furthermore, with the Windsor Framework set to resolve the issues of the Northern Ireland Protocol, there is fresh potential for the UK to secure associate membership of Horizon and other EU research programmes like Copernicus and Euratom. While we welcome the decision to extend the Horizon Europe Guarantee scheme until the issue of the UK’s participation can be resolved, reports in the press that the government may not decide to seek membership of Horizon after all are concerning. The programme’s strength lies in facilitating thousands of collaborative research links, inside and outside of Europe. This will be very difficult to replicate outside Horizon, and therefore ministers should pursue participation as a matter of urgency.
The government’s recent releases, including the Nurse Review and Future of Compute Review, provide a high-level vision, but the action that follows will be key and the government must up its ambition substantially. Such action should include:
Sovereign Capability in AI
Compute is one of those areas that might often be categorised as a specialist side issue – “invisible to the public” as the expert panel write in the Future of Compute Review – but it is absolutely essential infrastructure for the modern world. In essence, compute is the backbone of data, modelling and AI.
The changes the world has witnessed in AI over the past few months have shown that underinvesting in this field will be a strategic mistake. As it stands the UK has little more than 1 per cent of global compute capacity and none in the world’s top 25 systems. The Future of Compute Review describes this challenge well and its recommendations are structurally very strong. But given the pace of change, the government will need to deliver on recommendations much faster and with much more compute than the review asks for. The current recommendations would leave the UK unable to train a general-purpose AI system even to the level of ChatGPT until 2026, at a time of intense competition when other corporate and state actors will be racing much further ahead.
In particular, two things need to happen. First, the UK must deliver a genuinely world-leading supercomputer next year – more powerful than the two-exoflop supercomputer the US will build this year – in order to keep pace with leading AI labs. Second, it is suggested that the proposed UK AI Research Resource should provide at least 3,000 top-spec AI accelerators by the summer of 2023, which today would be equivalent to Nvidia H100 graphics processing units (GPUs). This will not be enough and it is our belief that three to four times this number is needed to ensure competitiveness with US labs and with the US's proposed National AI Research Resource. To achieve this, the UK government does not need to go it alone. It should strongly consider public-private partnerships with innovative UK firms already using large-scale cloud GPU services, which could help build the core infrastructure for a whole-of-nation effort in this space.
New Networks of Cutting-Edge Research Laboratories
While funding for Focused Research Organisations is a very welcome step in the right direction, much bigger changes are needed to embed new approaches to R&D in UK research culture.
As we highlighted in our report, the UK is falling behind the global pace of technology, even at the early stage, and is stuck in a habit of using old models of research funding.
Both our report and the Nurse Review of the R&D landscape highlighted the strong overdependence of the UK on a single model of research: universities. While they have significant strengths, they are not suited for all kinds of research, especially intensive development of new technology. As we highlighted in A New National Purpose, we should be investing in a network of new, disruptive institutional models for science and technology research, learning from examples like DeepMind, and also the Howard Hughes Medical Institute’s Janelia Research Campus, as recommended by Sir Paul Nurse, and benchmarked in funding to the best examples globally.
R&D Investment Boost
While there has been a substantial and welcome increase in public R&D investment, as we have previously noted a large fraction of this is consumed by bringing Horizon costs onto UK R&D budget-balance sheets (whereas it used to be financed through our EU contributions) and from a major boost to defence-research programmes. There is minimal uplift elsewhere.
As both our report and the Nurse Review highlighted, the UK is still very much in the middle of the pack for OECD investment in R&D and far behind the leading countries. Given the R&D budget is tiny compared to the overall public outlay but pays outsized returns, it makes sense to substantially increase it, looking to lead among OECD countries in direct public R&D spend by 2030 – and starting on that trajectory today. This is the only way in which to implement a serious reform agenda.
This new agenda requires a generational shift in policy thinking that is aligned to the scope of the change taking place in technology. We believe that this new national purpose will be critical for the future of Britain. The future will require investment. But the cost of inaction will far exceed the cost of making these necessary choices now.