Artificial Intelligence and Work

Download the report "Artificial Intelligence and Work"

Artificial intelligence – meaning the group of technologies that aim to carry out tasks traditionally assigned to human beings computationally – is central to current debate on social transformations. First and foremost, expected changes in the world of work give rise to two contrasting attitudes. Some people proclaim their optimism in the face of a technology that ensures productivity gains and is therefore a source of wealth, and which promises to do away with the most tedious tasks. Others make pessimistic prophecies on the inevitable disappearance of whole realms of activity and corresponding jobs. This being so, public debate is polarised in unproductive opposition as it fails to highlight factors of transformation or levers for action.

In order to help clarify such debate, Muriel Pénicaud, Minister of Labour, and Mounir Mahjoubi, Minister of State for the Digital Sector, entrusted France Stratégie with the mission of analysing the impacts of artificial intelligence (AI) on the world of work¹. This mission is complementary to the one that the Prime Minister entrusted to Member of Parliament Cédric Villani, which, given its wider scope, tackles questions of research, industrial policies and ethics. The goal is the same: educate to eradicate fantasy, but take stock of expected transformations while identifying appropriate public policies².

AI has been making spectacular progress over the last few years. Technologies resulting from recent research, such as machine learning and deep learning, have arrived from laboratories to carry out tasks that previously seemed impossible for machines to perform, such as image recognition, providing a satisfactory translation of a simple text, or winning a game of Go. Such technologies are already at work in our smartphones and constitute the framework of much of the pairing software already deployed in such areas as online advertising and profiling.

With the exception of a few specialised fields, AI is not yet much in evidence in most professions. This does not make its possibilities any the less significant, in particular for retail banking, transport and health – three sectors which are examined here in depth.

Artificial intelligence will undoubtedly be called upon to carry out complicated but repetitive or regularly performed tasks, which will obviously affect professions that include such tasks. But this transformation is not radically different from digitisation of the economy, which has been with us for some time and to which the banking, transport and health sectors have adapted more or less happily by modifying job content, training workers and developing new activities. Employees have also been adding to their skills for some time now in response to robotisation, in particular in industrial sectors, a fact that may guarantee their continued employment if it ensures growth of a company’s or sector’s activity. This is evidenced by the advanced robotisation of Germany’s automotive industry: in 2016, although it is one of the most robotised industries anywhere in the world, it had over 800,000 employees, 100,000 more than twenty years ago, as against 440,000 in France¹.

There can be no doubt that employees risk losing their autonomy by being subjected to increasingly insidious automated monitoring, with all the psychosocial risks involved. We are well aware of the controversy over working conditions in a number of warehousing facilities, where automated monitoring of employees is carried out by a system incorporating voice synthesis. Such systems may lead to greater fragmentation of tasks, performed with support from software tools.

None of these challenges is completely new, and improvement of working conditions is just as likely a hypothesis as alienation and work intensification. Everything depends on the way in which productivity gains enabled by artificial intelligence are shared and the choices made in task and team organisation.

Of course, factors other than technology have an impact on work. The behaviour of workers, customers and suppliers, qualification levels of a sector’s workers and possible tensions connected with lack of manpower, along with regulatory obligations, often play a determining role in the evolution of work.

What makes the present context different is that artificial intelligence is often based on a learning mechanism, with ongoing accumulation of data enabling continuous improvement of systems – to the point of creating, one day in five or ten years’ time or more, a real breakthrough in what it is technologically possible to do, depending on tasks concerned. Emblematic of such a future breakthrough is the promised advent of the autonomous vehicle. This revolution in mobility may eventually do away with the profession of driver, but at the same time it opens up a whole range of possible new professions in complementary activities. Construction, maintenance, fleet management and passenger assistance will remain, while recreational outings, logistics and straightforward professional travel will all benefit from lower costs and/or increased availability.

Spectacular progress is to be expected in all three fields examined in this Report: autonomous vehicles providing mobility, automated financial advisors in the form of chatbots, and medical assistants participating in monitoring health and wellbeing on a daily basis, as well as in prediagnosis and therapeutic proposals.

How many people are concerned in their everyday working lives? Potentially everybody, all the more so in that artificial intelligence tools are generic in nature, typically concerned with natural language processing and voice or image recognition. The 800,000 people in France who work as drivers are likely to see their work change radically as use of autonomous vehicles increases. The transformation will not necessarily be a sudden one, but will lead to orientation of work content to supervisory and reception tasks, or tasks that machines are unable to manage (such as finding the doorbell, for a delivery man).

All this may still seem a long way off, but it is already mobilising actors (innovators, longstanding professionals, customers and users alike), which also affects work transformation dynamics.

We must prepare ourselves for artificial intelligence, not because its advent is inevitable but because, in the society in which we live, technological possibilities open up new prospects for individuals, organisations and structures. There is nothing at all to be gained by lasting opposition to solutions that improve our fellow citizens’ health, provide access to safer and less expensive mobility, and provide cheaper financial services better adapted to consumers’ needs.

There is more than one way to go in such evolution, however, and this is where the public authorities must concentrate their efforts: setting a path that matches citizens’ social expectations, by defining appropriate controls on critical subjects (responsibility, security, etc.), and by accompanying evolutions that occur too quickly, so that the social and economic fabric adjusts naturally.

On the basis of the analysis it presents, the Report identifies three focuses to respond to the issues involved in artificial intelligence as far as the world of work is concerned:

• carrying out forward-planning work at industry and sector level on the potential of artificial intelligence, in order to ensure that stakeholders are fully informed and able to anticipate changes to come;
• seeing that workers receive training on the issues of tomorrow’s world: training highly qualified workers for production of AI, and workers who are fully aware of the technological, legal, economic and ethical issues involved in the use of tools based on artificial intelligence;
• improving schemes for safeguarding career paths in the sectors and subsectors that are set to be heavily impacted by the risk of automation.

Lastly, we should take care not to underestimate risks with regard to working conditions – loss of autonomy, work intensification, etc. – connected with the conditions in which AI tools are deployed in organisation of work.