Flux de publication de la direction générale du Trésor - Artificial-intelligenceFluxArticlesTag-Artificial-intelligenceCopyright 20262024-12-05T00:00:00+01:00/favicon.pngDirection générale du Trésorhttps://localhost/sitepublic/contact@dgtresor.gouv.fr00be60a5-dea5-437f-8a49-b44b5663ded2

The structure of the artificial intelligence value chain is a determinant of the innovation landscape. Tech giants play a major role in this landscape given their presence across the value chain. Although new entrants are challenging this status quo, it raises questions about economic efficiency, fair competition and sovereignty. France has major competitive advantages in this race, including data, a skilled workforce and an innovative research ecosystem.

2024-12-05T00:00:00+01:00<p>Artificial intelligence (AI) refers to various techniques enabling machines to simulate human intelligence. The AI value chain is divided into three main segments (see Chart below): (i) the inputs required to develop AI systems and services (computing power, data, specialised workforce); (ii) modelling, which includes the development of general-purpose AI models (foundation models) and specialised models; and (iii) the deployment of such models to end-users.</p> <p>Regarding AI system inputs, France has no major companies on par with the global leaders in the chip manufacturing and computing power rental markets. However, it benefits from a skilled workforce and a thriving innovation ecosystem.</p> <p>As for AI model development, a few French companies are emerging, but the segment is dominated by the incumbent Big Tech firms established prior to the advent of AI technology. These incumbents enjoy vertical integration thanks to their preferential access to upstream inputs and downstream distribution channels for their AI solutions (e.g. office suite software). They have also partnered with emerging AI firms to integrate AI production processes horizontally.</p> <p>The dominance of the AI market by a small number of large, already mature firms raises issues of economic efficiency, fair competition, and sovereignty, with the risk of limiting the economy-wide diffusion of AI-related added value and productivity gains.</p> <p>&nbsp;</p> <p style="text-align: center;"><img class="marge" src="/Articles/00be60a5-dea5-437f-8a49-b44b5663ded2/images/5302db42-7204-4e81-93a3-9cacab3184b6" alt="Visuel TE-354en" /></p>c29a855c-027b-44fa-b813-0cfce7679376

Artificial intelligence (AI) is a technology which has the potential to generate significant productivity gains that still go largely undetected at macroeconomic level, due to firms’ limited adoption of AI. Its impact on employment is more uncertain, although it could affect high-skill jobs to a greater extent than in the previous technological revolutions. Education, training and competition policy will play an essential role in helping everyone to reap the benefits of AI.

2024-04-02T00:00:00+02:00<p>Artificial intelligence (AI) refers to a set of techniques which enables machines to simulate human intelligence. Its development is a technological revolution which, much like with previous revolutions of this kind, could generate profound economic changes. While research to quantify the impact of AI is still in the exploratory stage, such work provides some preliminary insights.</p> <p>On a macroeconomic level, it is too early to empirically discern an impact on growth, but some initial microeconomic studies suggest that certain specific applications of AI have a significant positive impact on individual worker productivity. In a given job, these gains benefit the least productive workers the most, allowing them to catch up to their most productive peers. However, the impact of AI on business productivity has been found to be modest for the time being. This may be due to companies&rsquo; still limited and uneven adoption of AI, although there is more widespread adoption among large companies and digital firms.</p> <p>The theoretical impact of AI on employment is uncertain. In the short term, this impact will depend on the speed at which AI is deployed, the shift of certain occupations towards AI-complementary tasks and the reallocation of labour towards occupations in growing demand. Furthermore, initial empirical estimates indicate that the tasks and occupations impacted by AI will not be the same as those affected by previous technological revolutions. Skilled occupations are expected to be more impacted by AI due to its ability to perform abstract, non-routine tasks, whereas the previous waves of automation and computerisation had impacted unskilled occupations and mid-level occupations, respectively.</p> <p>These various findings point to the need to strengthen science curricula in primary and secondary education and AI curricula in higher education, to focus on continuing training for occupations affected by AI and to remove certain barriers to the diffusion of artificial intelligence, particularly by adapting competition policy to its particular qualities.</p> <p>&nbsp;</p> <p style="text-align: center;"><img class="marge" src="/Articles/c29a855c-027b-44fa-b813-0cfce7679376/images/27d96fb5-0e40-470e-a97d-b4284805409e" alt="Visuel TE-341en" /></p>