The development of AI — need for a paradigm shift
The labour market structure is changing all the time. One of the main disrupting trends in near memory has been the move away from manufacturing in developed countries. For example in the USA, the share of manufacturing jobs as a percentage of total employment dropped from 38% in 1950s to just 8% today. So far the economy and society as a whole has coped with this tremendous change remarkably well, despite some serious pockets of insufficient local adjustments creating the so called “Rust Belt” in the USA and its equivalents in the UK, Belgium, Sweden or the Czech Republic to name a few.
Today the use of (“narrow”) AI systems which carry out defined (structured) tasks is enabling automation of more tasks than ever. We are seeing an increase in the use of autonomous machines, due to the fact that they cost less, are more efficient, and can ensure a higher quality of work when compared to humans. This process of automation is taking place across industries such as retail, logistics and transport, and agriculture, but will also in the short term impact any jobs involving pattern recognition and/or repetitive (and often boring) tasks — radiologists, accountants, lawyers, sports reporters, customer support staff, analysts and managers, to name a few. According to McKinsey Global Institute between 400m and 800m people could see their jobs automated and will need to find new jobs by 2030.
As well as advances in narrow AI we are seeing increasing investment from individuals, companies, and governments in general AI (AGI). This general AI would not only be superior to humans at structured tasks, but in a wide variety of domains which involve general problem solving. Such AI will be able to perform complex mental tasks. Moreover, it is reasonable to expect that AGI will have the ability to learn and improve at any activity itself in a similar way as we humans improve with more practice (learning by doing).
Although there is not full agreement in the scientific community on whether AGI may be reached, many experts think it is possible. Therefore, it may be useful to start planning for a future where not only specialized (narrow AI) but also general AI systems are deployed, and think about the implications for public policy, regulations, ethics and other issues before the pace of change is too fast making it very challenging to keep up. Having considered the existing AI applications as well as the potential of AGI we are likely to see current trends intensify causing possibly a complete paradigm shift in work patterns which in turn may require new ways of adjusting in other areas.
Current trends intensify
- Shortening of supply chains, i.e. bringing production closer to end-users due to automation and new technologies such as 3D printing.
- Automation of not only tasks on the production line in manufacturing plants, but increasingly automation of less structured activities and back office tasks.
- Increasing power of large (globally) dominant companies which are able to commercially deploy new technologies at scale; accompanied (and reinforced) by a “talent drain” from other areas or sectors. (For example, it has recently been reported that the GCHQ, one of Britain’s leading intelligence agencies, is having trouble recruiting staff because Google and Facebook can pay more.)
- Empowerment of individuals who are able to take advantage of available technology and build new business models. This trend may increase as technology becomes cheaper and more widespread (democratization of technology). However, it seems reasonable to expect that not all humans will be able to take advantage of the technology to the same extent.
- Sharing economy / collaborative consumption — AI will increase efficiency of digital platforms and marketplaces which facilitate direct peer-to-peer transactions and may also increase effectiveness of services provided.
- Continuing rise of “gig economy,” as digital working becomes the norm, partially dissolving a strong bond between employees and employers. However, gig-workers will be continuously displaced by AI gig-workers (AI-gig-bots).
- Growth of inequalities between those who have access and are able to use technology and those who do not. However, it seems more important how big a slice of the AI co-generated “pie” is accessible to all citizens in absolute terms (eg: having enough to finance public service and other public spending) which would make everyone better off than in pre-AI age, than the relative inequalities. On the other hand, if the inequalities are not managed well we may witness growth of low-income groups.
- Technological changes lead to changes in the structure of society (family, elderly care, role of women). Some elements of this trend in particular in relation to new public services will be discussed below.
Overall, we may expect the emergence of “super-labour”: a labour defined by super high added value of human activity due to augmentation by AI. Apart from the ability to deploy AI, super-labour will be characterised by creativity and the ability to co-direct and supervise safe exploration of business opportunities together with perseverance in attaining defined goals. An example maybe that by using various online AI gig-workers (and maybe several human gig-workers), while leveraging AI to its maximum potential, at all aspects from product design to marketing and after-sales care, three people could create a new service and ensure its smooth delivery for which a medium size company would be needed today.
A way forward — (Possible paradigm shifts)
The above trends will lead to structural changes and the speed of change may be accelerating with the possibility of millions of jobs being lost globally in several years. Although there will be new jobs created in the private sector, e.g. in recreation, entertainment and applied arts, this profound change may still bring about a need to rapidly update the role of the state as well as private enterprises due to the need to maintain social stability. We can identify at least three main types of adjustment:
- The ways and means of provision of public goods and services will change.
- New public services will emerge.
- The financing of traditional and new public services will be reinvented.
This would mean nothing less than the overhaul of the welfare state. The establishment of the modern welfare state took place in the early 20th century as a response to the misery and degradation of life caused by the industrial revolution of the 19th century. Some historians also claim that the welfare state was created to placate workers, in order to prevent the rise of socialism. Early welfare systems were delivered by non-state providers, such as mutual and friendly societies, trade unions, insurance companies or churches. It is likely that apart from the increased use of technology in provision of existing public services, we will see a move towards non-state actors involved in welfare, due to the ability to provide better and cheaper services, but more importantly they will have the ability to adjust faster to these changing conditions (one such area being education, where AI may bring about disruption).
Among the strongest candidates for change in delivery means is also the care sector. With robots starting to take up menial tasks, such as delivering food and drugs in hospitals, and AI taking care of administration, it could free up time for care workers to focus on the human aspects of care. This could see previously unpaid jobs, like conversing with the elderly or reading to children, become paid ones. Patterns like this emerging in childcare and social and health care, would give carers more time (and also more carers) to focus on the important social and emotional aspects of their jobs. Other areas of new or rediscovered public services creation may include paid community-building initiatives and also more traditional public infrastructure projects such as parks, green spaces and leisure infrastructure.
Considering tax and welfare reform we may need to find ways to automatically adjust the tax base in order to generate the needed revenue as traditional sources of taxes, relying on taxing labour, may be at least temporarily reduced. For example, a higher rate of VAT for less human labour intensive (i.e. more AI intensive) production would generate both more tax revenue, as well as encourage companies to keep humans in employment in order to create a time buffer for structural adjustment. Furthermore, a higher income tax on those profiting most from AI technology could help speed up the implementation of a new welfare system, possibly through subsidised vouchers or Universal Basic Income (UBI), where certain services (potentially including newly paid services described above) are procured directly by citizens. However, there is an important issue to be taken into account when schemes such as UBI and accompanying tax reforms are considered: International competitiveness. With a possible introduction of UBI and tax adjustments, countries need to maintain productivity at a competitive level in the international markets.
Other paradigm shifts may be related directly to companies (or their shareholders). It may become their responsibility to derive new models to share AI co-generated wealth, either directly in the form of various grants or in the form of capital gains, e.g. creating new ownership structures favourable to individual investors such as cooperatives or through other special investment vehicles (which might generate income similar to UBI). Nevertheless, in the longer-term, if a state has the knowhow to build and deploy AGI together with the access to natural resources, the taxation may be reduced even to zero while being able to provide for their citizens.
Either way, with a potentially massive shift in work practices, there will be a need for a massive shift in labour related welfare calling for solutions such as new types of flexicurity where its active labour market policy element is targeted at both new jobs creation and match-making job seekers with new jobs.
In the longer term (but better sooner rather than later), we will need to overhaul our education systems. There is a lot being said and written about education however, one element seems rather neglected. We need to focus on how humans work and how we put in to use knowledge, skills and attitudes acquired in formal and informal education. It seems that most of us learn better through experience (learning-by-doing) than by analysing huge amounts of data. That makes us distinct as humans, at least when compared to narrow AI, and could help us develop competitive advantage in the area of work. In other words, we need to focus more on what a robot cannot do and/or where we outperform machines. We need to understand this better and then allocate time and space in the school curriculum to promote learning in that direction. We can expect that social and emotional skills will be part of the new curricula. The reason being not that the machines could not technically understand us and react to our needs (often even better that peer humans), but maybe it will be more socially acceptable that certain tasks are carried out by humans (or at least for those customers who can afford to pay a bit extra for a human touch). AI will also probably allow for new forms of human creativity to develop (the new AI sky will be the limit).
Some issues (to be addressed sooner than later)
Pace of change: It is important that governments and companies can keep up with the pace at which AI will transform work practices to avoid a situation similar to the above mentioned “Rust Belt syndrome” which saw whole regions suffer long-term economic downturn, urban decay and population loss. This seems to be a high price paid for deindustrialization; the workforce did not have the skills to transfer into different industries. The price to be paid in the case of AI deployment and resulting automation could be much higher. Indeed, apart from the structural changes discussed above (in particular any UBI type transfers need not to jeopardise international competitiveness) the mismatch between workforce skill set and future (automated) labour market is a crucial area that needs to be addressed now (see motivation and retraining below). However, there seems to be only a limited sense of urgency to act, e.g. in the recent poll by Gallup, the majority of Americans expect that AI will have a negative impact on the workforce, but at the same time the vast majority does not fear losing their own job to AI. We probably need to think more about how to provide factual information to the general public and what role governments and companies will have in this effort.
Motivation and retraining: We need to establish motivation mechanisms which will help speed up workforce retraining in order to increase the odds of securing new jobs for those who are under threat of job displacement. Although AI will replace jobs and at the same time create new jobs (similarly to previous waves of automation), new jobs will also be created in private and “new” public sectors as discussed above. In addition, for certain jobs it will not be cost-effective to automate them and they will remain for humans (e.g. bicycle repair in the short-run). For all these jobs skilled workers will be needed, at least before AGI allows also for their automation. The pace and potential of automation should be considered when recommending the direction of retraining. For example, simple coding is already now being automated so the growth of certain types of jobs may actually be rather short lived. On the other hand, jobs involving certain types of human interaction seem not to be easily automated in the early stages. In any case, AI can help us retrain people more efficiently inter alia through personalised learning.
Safety, ethics and governance: Making sure that all AIs are safe before being put into practice is paramount. AI should be tested in innovation labs and also in sandbox worlds, which it believes are real, in order for us to gain an understanding of how it will act in the real world, including the world of labour. These experimental worlds will allow us to fine-tune AI, avoid dangerous scenarios, and make sure the automated system acts in line with our ethics. Although we, humans, share a broad system of values there is no easy way to transfer them to machines. Other concerns are of course related to the abuse of AI by humans, be it for social control of others, or building autonomous weapons. Apart from ethical and safety issues there seems to be a crucial issue of governance of technology. AI, and AGI in particular, have highly disruptive potential. We need to be looking for mechanisms in order to ensure that the fruits of the technology are shared by the majority of people to benefit humanity, not only by a selected few. These mechanisms may, apart from direct regulation and fiscal instruments, include various self-regulation (see for example proposals of IEEE) and also new ways to protect and share intellectual property rights related to disruptive technologies. In the area of labour, future (accessible) AI could assist each individual by supplementing the very abilities which he or she lacks, but which are crucial for the success of the individual in the labor market and in society in general.
In the current interconnected world the spread and uptake of new AI technologies may be very fast. Those companies and individuals who will be able to leverage AI will be strengthening their competitive advantage, which may in turn bring about unprecedented levels of inequalities with the potential to disrupt current social order. Above we have proposed ideas which may be considered by key players, including companies and governments, in order to tackle the potential pitfalls related to technology development. Each of the ideas would obviously need to be elaborated on in detail and most of them would need to be made the subject of expert and public discussion. With the advances of AI we need to put more resources into understanding the nature of humans in the relation to the proliferation of autonomous intelligent systems and the role of work in our lives (a fundamental issue which we intentionally did not address in this text). Finally, we need to invest in developing governance models which will increase the probability that the benefits of A(G)I are shared by all.