Submitted, in April 2016, as written evidence to the Inquiry by the House of Commons Science and Technology Committee into Robotics and Artificial Intelligence also cited in the Inquiry’s main report.

---

Summary

1. The impact of robotics and artificial intelligence technologies (R&AI) on the UK work force and job market will be profound.  However, R&AI is not one technological advance but several meaning that the impacts can be expected to occur over several decades allowing time to adapt.
2. R&AI can help to mitigate the significant impacts of the ageing population on the UK society and economy, included shrinkage of the labour pool and rising costs for social care.
3. We should begin now to develop national and global technology governance systems that can regulate research and innovation in R&AI to ensure safety and that the benefits of these advances are widely shared.

Response to Inquiry

The implications of robotics and artificial intelligence on the future UK workforce and job market, and the Government’s preparation for the shift in the UK skills base and training that this may require.

• The implications of robotics and Artificial Intelligence (R&AI) for the future of employment in the UK are likely to profound and wide-ranging.  Nevertheless, these impacts can be expected to occur over several decades, allowing time to adapt.  The world has experienced three industrial revolutions as the result of the development of new technologies: (i) mechanical production through steam and water power, (ii) mass production through electricity, and (iii) computers and the first wave of industrial automation.  The 4th industrial revolution due to advances in artificial intelligence and next generation robotics is likely to have impacts at least as far-reaching as these previous three.  
• The effects on the UK workforce will be gradual because R&AI is not one technological advance but several—below I highlight four—however, each new breakthrough will directly impact on the job market by increasing the capability of machines to operate autonomously (i.e. without human supervision) or by improving their ability to understand the world (AI) or to act in the world (robotics).  
• One tipping point that is already upon us in the impact of advances in machine learning.  This has happened due to the coincidence of three factors: (i) massively increased computational power, (ii) the ability to amass and store very large data-sets (big data), (iii) improvements in learning algorithms.  The confluence of these three can be seen, for instance, in the victory of the Google DeepMind’s AlphaGo program against the world’s top Go player Lee Sedol. Whilst the significance of that is event is largely symbolic, there are clear real-world impacts that can be seen, for instance, in the latest real-time speech recognition technologies, and in the image processing tools that are transforming the ability of internet companies to exploit visual media.  More broadly, machine learning and probabilistic reasoning will lead to algorithms that replace human decision-makers in many areas of commerce as is already happening in financial trading. 
• A second tipping point, that is approaching rapidly, is for autonomous vehicles to be safer than those controlled by human drivers.  The core enabling technologies for this such as 3D sensing, visual scene analysis, and simultaneous localisation and mapping (SLAM) are already in place.  Current research is aimed resolving some of the special cases such as driving in poor weather.   Some of the major blocks to progress are legislative rather than technical, however, given the need for real-time decision-making and the direct threat to human life from errors, the role-out of these systems may be more gradual than expected.   Even so, many forms of transport could become driverless within a decade with substantial impact on both the UK economy and on jobs.  The Global Agenda Council estimated a widespread shift to driverless vehicle technology by 2026.
• A third tipping point will occur when robots develop the capacity for dextrous manipulation of deformable objects such as cloth and food.  This can be expected to have far-reaching impacts within the manufacturing sector.  Whilst robots are already very fast and precise they currently fall far behind human skill in manipulation with the consequence that whole sectors of manufacturing have resisted automation. In evolution, only primates (including, of course, humans), have succeeded in developing highly dextrous manipulation capabilities, and to achieve this level of mastery we have evolved particularly large brains. Manipulation has likewise proved to be a particularly thorny problem for robotics, and despite decades of effort, there is no sign that a breakthrough is imminent. Nevertheless, advances in sensing and machine learning technologies, together with the current degree of focus and investment in addressing this challenge could yield a breakthrough that could be rapidly rolled out since, in contrast to driverless vehicles, there are far fewer legal or safety obstacles to adoption.
• Finally, progress in AI towards systems that are able to reason about problems described in natural language and based on background knowledge of human society and culture will profoundly impact many white-collar jobs.   AI systems such as IBM Watson are already showing substantial progress in this direction, although there remains a gap between the capability of these systems to respond to a query with a correct answer, and for these systems to fully understand the meaning of the question or the significance of the answer.  We could characterise such systems as having a relatively shallow understanding in comparison to human experts. Nevertheless, AI systems of this kind look set to revolutionise access to knowledge in areas such as analysis of financial markets, medicine and law.  This will have a democratising effect (since access to knowledge will become easier and cheaper), whilst threatening the earning power of the professions that have previously been the gate-keepers to know-how in these domains.

Implications

• Whilst R&AI technologies will create jobs in the near-term, they will eventually take over much of the work currently performed by people leading to a net loss of jobs, or to a reduction in the amount of time people spend working.  A key benefit of R&AI technologies will be a general rise in standards of living, moreover, people will do less dreary, repetitive or unfulfilling work with the possibility of replacing this with activity that is directed towards family, social life, education, arts, and self-realisation.  
• Across Europe, we are moving from a situation where we have four working people for every one potential dependent (someone either over sixty-five or below fifteen) to one where, by 2060, that ratio could be two-to-one.  R&AI can help mitigate some of the most significant effects of this changing demographic, including labour shortages due to the retirement of older skilled workers, and the capacity of our health and social care services to support an increasing proportion of the population that are older and frail.  A new generation of safe and human-friendly robots could assist in extending the active independent lives of older UK citizens both at work and at home.  At work, robotic technologies can reduce the physical toll of vigorous or repetitive work allowing older workers to remain productive.  Assistive robots can also help compensate for the anticipated shortage in nurses and residential care workers by helping professional carers in their work, and by reducing work-related injuries, such as chronic back-strain, that otherwise lead some experienced carers to retire early.  Working with teams of assistive robots, the role of a professional carer could become more skilled and respected, less physical, less routine, and more focused on the people being cared for. 
• Concern about loss of jobs in the next decade or so should be tempered by recognition that people are very adaptable, and, in many professions will be hard to replace.  Compared to today’s robots, people are very safe, highly situation-aware, flexible, supremely energy efficient, strong (for our size and build), robust, expert at understanding language and the nuances of social situations, good at anticipation, able to manoeuvre in tight spaces.  Robots are far from matching humans on any of these measures.  For this reason, jobs that were expected to be automated decades ago, for instance that of the restaurant waiter, have proved remarkably resilient to automation.  Current projections of the loss of jobs to automation are largely based on highly simplified measures of the work that people do, and of how easily it could be automated, and should be approached with a large pinch of salt.
• Nevertheless, we are likely to see R&AI technologies entering into many more fields of work, but they will do so, largely, as part of human-machine teams.  Indeed, humans that are enhanced with R&AI technologies will usually outperform machines.   This is true, for instance, in the world of chess, where AIs outstripped grand masters in the 1990s, and yet the best chess players in the world today are not computers but human-computer teams.
• UK businesses that embrace R&AI will increase productivity and drive down costs benefiting the wider economy.   However, the societal benefits of R&AI could be put at risk by increasing wealth inequality. The modern notion of a market-driven economy is grounded on the ability of citizens to trade their labour in order to buy goods and services.  As machines do more of the work, households may lose purchasing power and without paying customers markets may become unsustainable.  As forecast by J. Rifkin in 1995, we may be approaching what has been termed a “post-market era”.  Governments, including the UK, will need to make substantive changes to taxation and social security regimes to address these challenges, or to consider more root-and-branch approaches such as a switch to universal income.  
• Education needs to prepare people for the jobs of the future rather than those of today meaning more access to R&AI technologies in the classroom.  As a society we will also need to educate people for lives that revolve less around work, and help them to find meaning and fulfilment in a world in which machines will be smarter than people.

The extent to which social and economic opportunities provided by emerging autonomous systems and artificial intelligence technologies are being exploited to deliver benefits to the UK.

• While the UK has a strong and agile SME base in R&AI, we lack large tech/AI companies that can compare to Apple, Microsoft, Samsung, Facebook, AliBaba etc. Large UK companies that are linked to the sector are primarily end-users rather than immediate developers of these technologies (with notable exceptions such as BAE Systems, Ocado, and Dyson). This is creating issues for exploitation of intellectual property generated by UK Universities even though we are recognised as having many world-leading research groups. 
• In the domain of robotics it seems likely that the future economic “dragons” have yet to appear. The current leaders in industrial robotics (including European companies such as ABB and KUKA) may not be especially well-placed to exploit what will be the far larger markets created by service robots. Internet/AI companies may succeed in transferring their technologies into robotics but this will require changing business models.  There is space for UK companies to have impact here.
• As with the mobile phone industry, there is likely to be a split between businesses that build platforms (robots) and those that provide control systems (AI), although the separation between these two may be less clean, than it is for smartphones and tablets, since robots will have more varied and bespoke physical configurations. 

Implications

• The UK is internationally competitive in research in R&AI and punches above its weight within Europe.  By learning lessons from the internet revolution, the UK does have the possibility to build companies that could compete internationally in R&AI but time is critical as the markets are emerging now. 

The extent to which the funding, research and innovation landscape facilitates the UK maintaining a position at the forefront of these technologies, and what measures the Government should take to assist further in these areas.

• The UK government has invested in robotics as one of the “eight great technologies” but the amounts committed to date are not yet sufficient to create truly transformational change. Sustained investment in people, in addition to the support for capital equipment, is needed.  Regional schemes, such as the Northern Powerhouse, could help translate research into innovation by encouraging the growth of hi-tech incubators near to major Universities.  InnovateUK also has a critical role to play in nurturing new R&AI businesses and should be appropriately supported to do so.
• The Alan Turing Institute is an important step in consolidating and advancing the UK’s position in AI and machine learning, creating a significant number of new opportunities for postgraduate and postdoctoral research.  A similar initiative in robotics should be put in place. 
• UK Universities are at the forefront of training future generations of R&AI engineers but the benefits are partly accruing to other countries due to a preponderance of overseas students doing these courses (partly due to a lack of home-grown talent motivated to do further study), and limits on graduates from overseas working in the UK after they complete their studies.  International students in R&AI help to maintain the UK’s competitive position in research, and to retain more of them in UK industry would provide a significant boost to the sector.
• At school level, the changes in GCSE computing (to emphasise actual programming experience) are a useful step, however, more could be done to promote learning about R&AI in schools, especially as experience with robots can inspire children to take an interest in the underlying technologies in computing and electronics and in STEM subjects more broadly.
• Apprenticeship programmes are needed to train the people who will install and service the very large numbers of service robots that we can expect to see deployed.

The social, legal and ethical issues raised by developments in robotics and artificial intelligence technologies, and how they should be addressed.

• It is useful to distinguish the ethical, legal and societal issues we are already facing in relation to R&AI from those that we may face in the future.  We already face issues about the use of robots as weapons (for instance, whether they should be able to make decision about use of lethal force), impacts of R&AI on privacy, and impacts on human autonomy (AIs making decisions on our behalf).  In the near-future we are likely to see an increase in R&AI in health and social care, an area where there are potentially many benefits but also growing concerns, some perhaps misplaced, about the replacement of human care by machine care. 
• In the longer-term, there is the question of whether R&AIs could achieve a level of general intelligence and autonomy where we would need to see them as other beings (potentially with rights), where people will form personal relationships with R&AIs, and where there is ultimately a risk of humanity being surpassed by R&AI as the leading intelligence on this planet.  While these outcomes may seem like science fiction, they are at least theoretically possible, and the UK population is already expressing concern. 
• One important consequent of these technologies is that they will not only change how we live our lives, they may also have a profound influence on our understanding of the human condition.  For instance, the development of artificial systems that can communicate, reason, engage with us socially, etc., will likely force a rethink of some of our core cultural ideas about life, free will, the nature of the human self, and so on.
• We have learned from studying climate change that human technologies can have potentially irreversible world-changing effects.  It is not over-dramatizing to say that long-term impacts of R&AI technologies could be as significant. 

Implications

• Concern about R&AI appears to have become something of a lightening-rod for wider fears about technology meaning that one of the major obstacles to the development of social and economic opportunities in the UK is likely to be public opinion.
• There is a need to address public concerns, and to ensure that developments in R&AI are safe and for the long-term benefit of the wider community. This can be achieved through better science communication and through improved governance of R&AI research and innovation.
• Research in R&AI is worldwide, and is conducted by many different types of organisations including traditional bodies such as universities and corporations through to newer and more informal players such as online and open-source “maker” communities.  Governance and regulation will need to be international if it is to be effective and not simply promote competitive advantages for less regulated countries.  The UK can provide global leadership in meeting this challenge.

---