In July 2004, the Government of the day published its Science and Innovation Investment Framework 2004-2014, in which the headline goal was to raise the level of UK investment in R&D from the prevailing value of 1.9% of GDP to 2.5% over the next decade. Today, the latest thinking is laid out in the current Government’s Industrial Strategy, which sets the headline goal of raising the level of investment in R&D from the current level of 1.7% of GDP to 2.4% over the next decade. Has time stood still?
There is currently widespread focus on the UK economy’s relative lack of productivity, in comparison with, say, the US, Germany and France. All of these countries enjoy much higher levels of productivity than the UK and have also been able to sustain much higher levels of R&D investment as a percentage of GDP. This raises important questions for science policy, including how public investment can lever increased levels of investment from the private sector.
Thursday 26th April saw the publication of two separate pieces of work designed to help address these challenges: the Government’s AI Sector Deal, and the Bond Review of Knowledge Exchange in the Mathematical Sciences. Both documents make eloquent cases for increased public investment, first in mathematics (the Bond Review) and then in data science and AI (the AI Sector Deal). They both recognise that the productivity challenge can be addressed through boosting R&D investment, and make recommendations accordingly.
However, when read side-by-side, these documents reflect the fragmentation in the UK’s research base that has dogged for decades attempts to increase R&D investment. The AI Sector Deal makes no mention of mathematics, except for the usual statements about education in STEM subjects. The Bond Review, focussing on the value of research rather than education, argues that: ‘Machine learning, artificial intelligence (AI) and data science are dependent on mathematics to find patterns in complex datasets’. It seems pure coincidence that these reports were published on the same day.
So is mathematical research important for the development of AI? I would say yes, and that Bond is clearly right on this point. But does the AI Sector Deal recognise this? It’s hard to tell, and if so then it’s well hidden. No wonder that Lord Stern, in his preface to the Bond Review, observes about mathematics that ‘Yet its role is not well understood, we are not using it as well as we could and should, and we are investing too little’. This statement applies widely, but the publication of the AI Sector Deal gives it a particularly sharp focus.
Perhaps the difficulty in coordinating contributions from different parts of the science base is to do with speed of change. Modern economies have shifted rapidly from manufacturing, to services, and now to personalised customer experience. The mathematics that we need has evolved in step, from the mathematics of physical processes (manufacturing) to the mathematics of networks and stochastic processes (services) and now the mathematics of data (customer experience). There is no shortage of possible opportunities, but we must be clear about what the opportunities are and ensure that businesses can exploit them, with the support of clear public policy and action.
In short, the challenge is in turning opportunity into reality. The gap in Technology Readiness Levels (TRLs)1 between business and academia remains too large, which leads to many opportunities to exploit mathematical insights being lost before they are even recognised. This is despite the great strides that many university mathematics departments have made in developing long-term initiatives in knowledge exchange. It does not help that mathematics is rarely adequately represented at the ‘top table’ when major public science and technology programmes are being designed; this is a difficulty that runs through the current Industrial Strategy. One hopes that the new AI Council (part of the implementation of the Sector Deal) will include a high-level mathematical voice.
The pace of technological change is so great that most businesses are not equipped to address the challenges for themselves. These are the businesses that are identified as ‘Practitioners’ or ‘Aspirationals’ in IBM’s recent global C-suite study2. One might think that universities would find plenty of opportunity in such an environment, but in fact the role of universities as a source of new business ideas faces increasing competition. In an era when companies focus on customer experience, they often look to their customers as co-creators of new products and services, rather than third parties. As a result, especially if the problems of fragmentation in the science base and public policy cannot be addressed, there is a real possibility that over the next few years the contribution of universities to innovation will shift away from the output of ideas towards the output of people. There will be premium on those who can think effectively outside the box, and mathematics departments are full of such people. Whether public policy should concentrate on ideas or on people, as a nation we should be making much more of our mathematical resources.
Returning to R&D, it is important to remember that the private sector contributes more than two-thirds of overall investment in the UK. Therefore, if we believe that R&D investment leads to improved productivity, then it is paramount that we shape public investment to stimulate more business investment. If the right choices are not made, 2.5% of GDP will remain a long way off. We can then look forward to a continuing productivity gap and a further government strategy paper in around 2030, aiming to increase R&D investment to 2.5% of GDP by 2040. If mathematics is properly harnessed by business and government alike, then we may yet stand a fighting chance.
1 Technology Readiness Levels are a 9-point scale developed by NASA as a means of tracking technology development, from observation of basic scientific principles (Level 1), through successful mission deployment (Level 9).
2 Incumbents Strike Back: Insight from the Global C-Suite Study, IBM Institute for Business Value, 2017.
