The victory of the machine on the human specialists of the game of go, first tests of autonomous vehicles, multiplication of the personal assistants such as Siri (Apple) or Cortana (Microsoft) … The artificial intelligence has become in recent years very media. It is an area in which it raises many hopes: that of business transformation. The deployment of AI, like that of other technological developments before it, will probably result, in terms of employment, in a creative destruction.
Artificial intelligence is in the air
The issue of Artificial Intelligence (AI) is at the heart of society, as evidenced by numerous reports and studies in progress, or the mission entrusted by the Prime Minister to Cédric Villani.
In 2016, the CBInsights platform identified more than 1,600 start-ups specializing in the field of AI. A recent prospective analysis of Statista also reveals that the rapidly growing AI applications market is expected to be worth more than $ 30 million in 2025.
Beyond the question of the reliability of forecasts made by research and consulting firms, we can question the impact that the IA will have on the transformation of trades, as well as the associated societal changes. Indeed, significant financial resources are committed: the European Commission, in particular, has planned € 1.5 billion to develop research on AI between 2018 and 2020.
No, your child will not become (next) terminator
AI was a concept that appeared in 1956, but its genealogy dates back to the early work on formal neural networks in the 1940s. These had already been set to replicate the cognitive functions of the human brain, thanks to advances in the computer. The AI concept is based on several disciplines dealing with natural language processing, logical reasoning, cognitive science, and so on.
One of the key principles of AI is learning from data collected through different platforms and computing devices. The increasingly massive collection of data (in connection with big data) coupled with the incredible increase in computing capabilities (Moore’s famous law) and advances in algorithmics explains the progress made by the AI these last years. We can not dissociate the AI from the advances made in machine learning techniques and in particular deep and machine learning.
On the other hand, the simplistic temptation to present technologies developed from AI as simply allowing automated processing of repetitive and tedious tasks must be avoided. AI is indeed capable of complex actions, such as driving autonomous vehicles. But we should not imagine that the AI could evolve into an intelligent machine, which would take its autonomy vis-à-vis its creator-programmer. At least for now. Indeed, the IA is the result of a software architecture (the type of neural networks, number of layers, learning methods, etc.) resulting from a human logic.
There remains a huge gap to be filled before arriving at a “strong” AI, which would be comparable to a human intelligence, especially in terms of the ability to understand contexts of interactions between individuals and/or objects.
The AI an event vector of job-creating destruction?
As with each major technological change (Internet, robotization, etc.), with the emergence of the AI raises the question of the consequences in terms of jobs and qualifications. On these points, the work currently available provides contradictory results but nevertheless, the threat to employment still seems very limited, as underlines the France Strategy study.
AI, which should automate some repetitive tasks, will transform many trades: it will require a strong evolution of qualifications (in maintenance, assisted creation etc.) and skills. And not just computer skills: in a recent study, McKinsey determined the impacts of AI on key competencies in different sectors.
AI-related technologies, which rely on massive data processing through connected devices, could reinforce the needs of newly created trades (data scientists, AI programmers, etc.). They could also bring out new ones: service delivery platform managers, artificial intelligence personality designers, knowledge engineers, etc.
Many trades are to imagine to allow integration between business needs, which will become increasingly complex and decompartmentalized because of the possibilities allowed by the AI, and the configuration of these technologies. There could be “AI integrators” in organizations that would be cross-cutting project leaders in the development, integration, and maintenance of AI systems.
Beyond this phenomenon, now classic, “job-creating destruction”, raises the issue of awareness and training of employees in positions related to AI. Training in technical, but also legal, social, economic and ethical issues. The risks of worsening inequalities or social polarization are indeed very real and have been identified by several studies.
The content of the training offer of higher education (engineering schools, management, legal training, etc.) will have to evolve rapidly, to integrate these dimensions.
Social acceptability of the AI, the real question?
The issue of AI is part of a context and trend of digitization of economic activities. In recent years, business transformation has been closely associated with the availability of data, the quality and capacity of digital infrastructures and, consequently, the trust of users (individuals, businesses, administrations, etc.).
The recent scandals concerning the use of personal data by companies show that ethical and even moral issues become a major issue. The development of AI risks considerably accentuating these issues and raises many questions: what is the responsibility of the designer of the AI? What precautions should be taken when AI integrates into humanoid forms or robots? What should be the degree of transparency in human-machine interactions? How to protect data shared between various AI interfaces? How far do you learn AI? Etc.
More broadly, there is the central question of the social acceptability of AI for employees, users and citizens alike. How far are we willing to go in the automation of tasks or services, knowing that there is a phenomenon of learning and acculturation? Technologies are sometimes ahead of uses and it has become difficult to understand how users will appropriate them.
For example, in the case of autonomous vehicles, manufacturers are close to reaching the last level of autonomy, namely level 5. However, studies show that few drivers of vehicles are willing to fully trust a fully autonomous vehicle), let alone when the vehicle has to carry children. To change this level of trust, several factors can play such as the legal context (who is responsible in the event of an accident between two vehicles: the owner of the vehicle? The creator of the algorithm? The constructor?) Or the acculturation the automation of other modes of transportation, such as the metro or bus. The conversation of course, before society can reap the full benefits of AI applications, many questions will have to be answered …