Linked articles

The Défenseure des droits calls for vigilance regarding the use of algorithms in public services

Social benefits, tax calculation, access to higher education: more and more administrative procedures are now partially or fully automated. What remains of human intervention in decision-making? What impact do these new practices have on users' rights?In a report published on November 13, 2024, the Défenseure des droits, responsible for defending the rights and freedoms of users of public services, examines the growing use of algorithms or artificial intelligence (AI) systems in French public services, and particularly in central government."Despite gains in decision-making efficiency ", the Défenseure des droits warns against " infringements of users' fundamental rights".Massive algorithmization of public servicesIn this report, the Human Rights Defender examines the effectiveness of two guarantees that are particularly important in ensuring respect for these rights: human intervention in decision-making and control of systems, and the requirement for transparency towards users.In order to automate, improve or accelerate certain procedures, the administration is increasingly resorting to algorithms or AI systems. Individual administrative decisions can be partially automated - if human intervention is involved - or fully automated.For the Défenseure des droits, this development brings both progress and risks for the fundamental rights of users. On the positive side, if algorithms and AI systems are appropriated and used responsibly and in accordance with the law, these developments can lead to more efficient and better-informed decision-making through the processing of a multitude of data, or help to limit arbitrariness and achieve equal treatment, as human agents are not always neutral and their decisions can be discriminatory. On the other hand, algorithmic systems also pose major risks to rights and freedoms, as the Institution has emphasized in two reports, one in 2020, on the risk ofautomated discrimination induced by algorithms, and the other in 2021, on the imperative need to respect fundamental rights in the use of biometric technologies.Partially automated" decisions: human intervention as a guarantee of user rightsWhen an administrative decision is said to be "partially automated", a public official must perform a positive, concrete and significant action based on or alongside the result generated by the algorithm, in the decision-making process.The Défenseure des droits notes, particularly on the basis of the complaints it receives, that this intervention is sometimes non-existent - as in the case of the Affelnet high school placement procedure, or Parcoursup - and sometimes inconsistent or biased, when the people involved in individual decision-making tend to endorse the results produced by the system without questioning them.In view of these limitations, and their impact on the rights of public service users, the Human Rights Defender recommends the introduction of mandatory criteria and operating procedures to define more precisely the nature of the human intervention required.Risks of infringement of fundamental rightsAlthough user complaints are still few and far between, the Défenseure des droits has identified a"systematic" problem of trust between users and the administration. In this respect, it points out that the transparency of public action must be seen as a "prerequisite for combating possible errors, abuses and discrimination", in the spirit of Article 15 of the Declaration of the Rights of Man and of the Citizen. However, information obligations are sometimes poorly respected, due to the opacity of procedures carried out by or with the help of algorithms. The report also points to the risk of automated discrimination. While in France there is a ban on making individual administrative decisions on the basis of sensitive data (such as economic situation, gender, place of residence...), the General Data Protection Regulation (GDPR ) remains less exhaustive and restrictive in this area. Examples, such as the tool developed by the Austrian national employment agency in 2019, show that this risk of automated discrimination has been able to materialize, including in the public sector.The duty of transparencyWhen a decision is based on algorithmic processing, a public service must provide a certain amount of information not only to the user concerned, but also to the general public. This legal requirement for transparency, which derives from a constitutional principle, must enable certain decisions to be understood, so that they can be debated or even contested.What are our recommendations?Faced with the risk of infringement of users' fundamental rights, the Défenseure des droits lists several recommendations:specify the framework (criteria and modus operandi) for human intervention by a civil servant when a public service decision is made with the support of algorithmic or AI systems; ensure that in-house skills are available and that staff are fully conversant with the operation of these systems; introduce penalties for non-compliance with users' right to information;enshrine a right to explanation of partially or fully automated individual administrative decisions; support research to understand the effects of these systems and fuel public debate; encourage the involvement of associations in this area, question objectives and reorient the system in line with user feedback. 

• Data

• Inclusion

• Citizenship

Published on 26/11/24



[Feature] The impact of artificial intelligence on work and employment: a new challenge for social dialogue

ForewordThe future of work is one of the five major themes of the Summitfor Action on Artificial Intelligence, to be held in Paris on February 10 and 11, 2025.The opening of ChatGPT to the public in November 2022 has rekindled long-standing concerns and debates about the potentially massive effects of artificial intelligence (AI) on employment. In March 2023, Goldman Sachs estimated that 300 million jobs worldwide could be destroyed by the irruption of this technology. In April 2024, consulting firm McKinsey estimated that by 2030, more than 30% of the hours worked in Europe and the USA could be automated by generative artificial intelligence (GAI).Meanwhile, the Artificial Intelligence Commission, in its March 2024 report, drew a more nuanced assessment of the effects of AI on employment in companies that adopt AI, as it replaces tasks, not jobs. " However, we must prepare for the disappearance of certain professions and anticipate the transformation of all others (...) The consequences of AI on the quality of life at work will depend on our collective choices and the quality of social dialogue with regard to it" concludedthe commission's experts.One thing is certain: " AI will transform the labor market", concludes the French Economic, Social and Environmental Council (CESE) in an opinion issued ahead of the Summit. In particular, the CESE points to the speed at which AI is being deployed in companies in the absence of consultation. " AI needs to be questioned, thought out and integrated as early as possible, as it can be a lever for improving working conditions (management, work investment)" concludes the CESE at the end of a controversy analysis. "A balance needs to be struck between the functions performed, the conditions under which they are carried out and the meaning given to the work. It is a major challenge that companies, administrations and professional sectors must seize upon".The LaborIA action-research laboratory, set up by the French Ministry of Labor and lnria, for its part, is publishing groundbreaking results on human-machine interactions and the challenges of appropriating AI in the world of work.TheInstitut de Recherches Économiques et Sociales (IRES) andANACT (Agence nationale pour l'amélioration des conditions de travail), for their part, worked for 18 months with some 50 participants from the worlds of labor and management, business and government, to develop an " operational toolbox designed to foster technological social dialogue around AI". The uncertain effects of AI on employment and productivity"Work on quantifying the effects of AI on growth and employment is still exploratory," observed the French Treasury Department in April 2024, in a summary note: " The effects of AI measured on business productivity are modest for the moment. This can be explained by a still limited and uneven adoption within companies, stronger for large and digital companies (...) The theoretical effects of AI on employment are uncertain."According to the International Monetary Fund (IMF), 60% of jobs in advanced economies could present a high degree of exposure to AI: 27% of jobs would be highly complementary to it, and therefore most likely to benefit from AI, while it could substitute for 33% of jobs. According to the International Labour Organization (ILO) , in developed countries, the number of jobs with the potential to be enhanced by AI (13.4%) is much higher than those with the potential to be replaced by AI (5.1%). According to other estimates, focusing more specifically on the arrival of foundation models, while 80% of US workers could see at least 10% of their tasks replaced, only 19% could see this share rise to at least 50%, and would therefore face a significant risk of substitution. " However, these results must be interpreted with caution," stresses the Treasury Department (in the same summary note). "Indeed, the approach adopted takes into account neither the progression curve of AI, nor the evolution of its development costs for companies, whereas these two elements largely determine the long-term effects of a technology on employment".The Artificial Intelligence Commission, in March 2024, pointed to a rather positive effect of AI on employment in companies that adopt AI, "because AI replaces tasks, not jobs. In 19 out of 20 jobs, there are tasks that AI cannot perform. Jobs that can be directly replaced by AI would therefore represent only 5% of jobs in a country like France. On the other hand, the spread of AI will create jobs, not only in new professions, but also in old ones. All in all, some sectors or fields could experience net job losses, which need to be supported by the public authorities, but this does not imply that AI will have a negative effect on national employment in France."To draw up this generally reassuring assessment, the experts commissioned by the President of the Republic relied, in particular, on a survey carried out annually by Insee, which studies the effects of AI adoption by companies in France. According to this study, total employment in companies that have adopted AI is increasing more than in companies that have not, whereas these two groups were following a similar previous trend. The effect results mainly from the creation of new jobs, rather than a greater retention of existing jobs . "We also find that there are no differential effects on male versus female jobs: AI adoption has comparable effects for male and female employment. However, this effect of AI on total employment is not uniform from one trade to another. In particular, certain levels within the company or certain trades are likely to experience net job reductions. In fact, companies that adopt AI for administrative management or marketing will see their employment in "intermediate administrative and sales professions decline. " Opportunities and risks of artificial intelligence for companies and employees"The opportunities and risks of artificial intelligence (AI) for the world of work are numerous. For the moment, their scientific assessment remains limited", observe Flore Barcellini and Moustafa Zouinar (Centre de recherche sur le travail et le développement- CRTD, Cnam) and Nathalie Greenan (Centre d'études de l'emploi et du travail-CEET).The promises associated with AI are essentially based on three assumptions, observe the authors."On the work side, AI is supposed to increase human capabilities in carrying out certain tasks, help reduce errors by making tasks more reliable, or relieve humans of tasks perceived as 'non-value-added', particularly when they are repetitive or arduous."AI also opens up opportunities in terms of employment and production structure. "The development of the technology, its implementation and the product or service innovation it stimulates should generate new jobs. Some even see Industry 4.0 including AI as an opportunity to reindustrialize the European economy."Last but not least, AI would also provide a partial response to labor shortage issues in the context of an aging population. This vision is reflected in the 5.0 society project promoted by the Japanese government.These hypotheses, however, "have been the subject of few solid scientific evaluations in real-life contexts. Moreover, current AI is mainly "weak" AI, specialized in specific tasks, and not "strong" or "general" AI capable of fully replacing human intelligence in all fields".Beyond promises, major risksthe partial or total replacement of humans, implying the loss of their expertise and of what gives meaning to their work: " this manifests itself through the increasing automation of tasks, including those considered 'cognitive' and based on the experience and expertise of professionals (e.g. medical diagnostic tasks)".increasing subordination to algorithmic systems and to the constraints of work organization: "this can be seen in the development of algorithmic management or the rigidification of work organization rules linked to the introduction of AI".the mirage of collaboration between humans and AIS: " human-AIS collaboration refers to the idea that humans and AIS would not be substitutes, but could work in collaboration, enhancing each other's capabilities - AIS potentially enhancing human capabilities, and humans 'training' AIS with massive structured data, providing explanations of its actions and monitoring its proper functioning"."Critical analysis of recent empirical results obtained by research in economics and ergonomics invites us to question a form of "solutionism", which leads us to see technologies as "simple, off-the-shelf remedies" for complex problems such as health, performance or cost reduction issues" conclude the authors.ice.s.To encourage the emergence of " virtuous scenarios of AI use ", the authors point to two central elements:AIS uses must be part of sustainable forms of work " that combine organizational performance with the preservation of workers' health and the development of their skills throughout their working lives"." The introduction of AI should improve the quality of the knowledge produced and facilitate the exploration of new knowledge on the production process of companies and the work of professionals". Cautious appropriation of generative AI by businessesTo date, companies have opted for a cautious approach to generative AI, based on successive experiments.According to Bpifrance's latest business survey, based on feedback from nearly 5,000 companies, 31% of VSEs and SMEs were using generative artificial intelligence (GAI) by the end of 2024, a proportion that has doubled in one year. In particular, 8% use it regularly (3% at the end of 2023). Conversely, 69% do not use it at all, although 19% plan to do so in the near future. Roughly half of all managers are therefore reluctant to use EMI, although this proportion has fallen sharply over the past year (72% at the end of 2023). The main use of EMI identified, which has risen sharply over the past year, is the generation of written content, requested by 68% of VSE/SMEs using EMI (versus 54% at the end of 2023). The second most common use is the search, collection and analysis of data or information (57%). Around 1/3 use it for translation (36%) and the generation of various visual content (31%). VSE-SME managers who do not use AGI explain it above all by the lack of usage identification within their company (for 70% of them). A quarter cite a lack of expertise on the part of their employees or themselves, and 13% fear misuse of the tools (sharing of confidential company data, lack of verification of content produced by the applications, etc.).According to a study by the Adecco Group, published in spring 2024, based on a survey of 2,000 business leaders in nine countries including France, the majority feel overwhelmed by the AI revolution. 57% of executives doubt their own management team's ability to grasp the "risks and opportunities" associated with AI. Only 43% of this group said they had formal training programs in place to improve AI skills, while only 50% said they provided guidance to their teams on how to use it at work. Although 66% of executives recognize the impact of AI in their industry, only 11% feel they have made progress in digitizing their business in recent years (versus 9% of French executives and 14% of American executives). 66% of managers (64% in France) plan to recruit externally to cope with this revolution. This trend towards outsourcing skills is also confirmed for other professions such as data processing, for which 62% of managers plan to recruit experts from outside the company rather than favour internal recruitment. Only 34% of managers intend to train their staff in the use of AI. French executives stand out, however, by favoring the reorientation of their employees towards other positions in the event of AI impacting their jobs (51% of French executives versus 46% worldwide).Employees ahead of the game when it comes to generative AIThe Terra-Nova think tank's report on generative AI draws on the positive effects of generative AI already observed in the development of professional skills, vocational training and the enrichment of certain tasks, to identify the conditions under which this technology could become a lever for inclusion and improved quality of life at work (organization and working conditions, skills enhancement, career development). The report does, however, note a certain reluctance on the part of employers to embrace this technology. " There is talk of shadow GPT in companies, a phenomenon of invisible appropriation by employees, which shows an appetite or need far removed from the wait-and-see attitude or fear that employers often attribute to their teams. Nonetheless, this take-up of the tool is not always without risk for organizations ".A survey of over 14,000 workers in 14 countries in 2023 revealed that generative AI tools were penetrating the world of work without supervision. More than a quarter (28%) of the employees surveyed were using generative AI at work, and more than half without their employer's formal approval. Pointing to a lack of training, they noted that their employers' generative AI policies were either unclear or non-existent. It is therefore important, conclude the authors of the Terra-Nova report, to make generative AI " a subject of social dialogue within companies, by gradually making it a common subject of possible agreements such as the employment of seniors or disability, or by positioning it, when the time comes, as a subject of compulsory agreements such as teleworking or professional equality between men and women. In addition to supporting the career paths of employees exposed to the deployment of AI Gen, this social dialogue will also need to address the conditions for ethical management of the potential of this technology. At both company and branch level, it is also important that the social partners work together, within the framework of the Gestion des emplois et parcours professionnels (GEPP), to study the possibility of detecting trades that could be in decline or on the contrary in tension, that they work on the evolution of skills and that they consider the interest of setting up a joint observatory of the effects of AI on trades". CESE: "Co-constructing a new social dialogue is essential".Just a few days before the Summit for Action on Artificial Intelligence, as part of its global reflection on AI, the Economic, Social and Environmental Council (CESE) has published an opinion listing the conditions for AI to "bring both social and economic progress".This opinion is based on two observations:For employees, "fear of losing or changing their jobs, feeling employees are very wary of artificial intelligence (AI)".As far as employers are concerned, " the majority of companies are still in the AI learning phase, with empirical experiments and tests on certain tasks. The fact remains that AI is disrupting employment and work by inducing a complete change in organizations that can destabilize them.""Prospective studies on the effects of artificial intelligence have reached a number of limits ", observes the CESE: "there are major discrepancies in estimates of the number of jobs created or destroyed, the CSPs most affected, and expected productivity gains. It is essential to recognize that the decisions taken today in terms of employment, labor and training, as well as in terms of regulation, will determine the impact this technology will have for both employees and employers".The EESC used the controversy method to analyze the arguments for and against AI in work and employment. This method was applied to three major issuesWill AI increase inequality at work?Is AI a step forward for the organization of working conditions and occupational health?Will AI have a positive impact on employment?The CESE has identified the groups most specifically affected: administrative employees, who have a large proportion of tasks that can be automated; women, because they are numerous in exposed professions (accounting clerk, administrative agent, insurance executive); and young graduates, who generally acquire business skills by starting with repetitive tasks, which in future will be entrusted to AI.Social dialogue does little to address the arrival of AI in companiesSince 2017, just under one agreement in a thousand has made reference to AI. This situation is changing, however: between 2018 and 2023, the proportion of signed agreements that mention AI has increased 2.5-fold, particularly in the context of mandatory annual negotiations (NAO).For the CESE, the introduction of AI in companies must be carried out in a transparent way that reassures all stakeholders. Transformations in skills, jobs and work must be anticipated through constructive dialogue.The EESC's analysis of controversies highlights 9 points for attention, "themes necessary for a constructive dialogue between employees, their representatives and employers for the successful implementation of an artificial intelligence system (AIS)".the need for clarity in the company's approach at all levels: it is important to ask the right questions and to advocate transparency when introducing AI in a given department or business line, to anticipate the induced impacts, and to avoid the many biases amplified by the use of AI (gender, disability, skills...).consequences for job content: skills to be replaced, new skills to be acquired, training and support required;consequences for work organization and working hours ;monitoring the impact on physical and mental health (risk of intensification);conditions for preventing AI biases (gender, disability, age, and any type of stereotype), since AI works with data programmed by humans;the sharing of value and expected productivity gains (wages, working hours, etc.);data protection for employees and companies (avoid outsourced programming);environmental impact;access to this technology for businesses (especially SMEs). Dia-IA: an operational toolbox to foster technological social dialogue around AIFor 18 months, some 50 participants from the worlds of labor and management, business and government, worked on developing Dial-IA, "an operational toolbox designed to foster technological social dialogue around AI".This project, coordinated by the Institut de Recherches Économiques et Sociales (IRES) and co-financed by ANACT (Agence nationale pour l'amélioration des conditions de travail), involved four trade unions in particular: CFDT, CFE-CGC, FO Cadres and UGICT-CGT.The project resulted in a joint manifesto and a series of fact sheets designed to help stakeholders become acculturated:AI basics;Mapping the issues;AIS development situations.This toolbox reviews the various reasons why employees and employers are talking and negotiating about artificial intelligence.The toolbox also includes "levers to activate in order to implement a technological social dialogue approach", aimed primarily at employee representatives. This toolbox also offers "levers to activate in order to implement a technological social dialogue approach", aimed primarily at employee representatives.Read more : Workers and artificial intelligence: what kind of social dialogue? AI in the world of work: a survey and recommendations for "empowering AI"After two years of surveys, the LaborIA action-research laboratory, set up by the French Ministry of Labor and lnria, has published groundbreaking results on human-machine interactions and the challenges of appropriating AI in the world of work.After gathering the perceptions of stakeholders (decision-makers, designers, engineers and employees) in different types of organizations (private companies, government agencies, public institutions), the survey produces three types of results.The deployment of AIS in organizations is not the end point of innovation processes, following the phases of ideation, prototyping and experimentation. Human-machine interactions involve extended and uncertain learning periods. Workers must not only use AIS, but also engage in its maintenance, enhancement and supervision.Work that is often little or poorly recognized by organizations can lead to disengagement and the long-term failure of AI projectsThe failures and successes of AI projects also depend on conflicts of priority in the workplace. The survey reveals an opposition between a managerial logic of AI, promoted by designers/decision-makers, and a "logic of real work, specific to employees". The managerial logic sees AI as "a means of optimizing processes, reducing the risk of errors, improving performance and increasing work productivity".The deployment of AIS in organizations can have unexpected effects on work organization and management. Among these unexpected effects, a reconfiguration of professional roles and qualification frames of reference, a questioning of the role of middle managers, polarization of work, etc., are just a few examples. What's more, if AI changes work (and its organization), work also changes AI: "different modes of work organization - highly hierarchical/centralized or, on the contrary, leaving more room for autonomy - strongly influence the reception and appropriation of an AIS".At the end of this survey, LaborIA makes six recommendations.Read more: Deploying AI in the world of work: survey and recommendations for "empowering AI" The future of work on the agenda at the Artificial Intelligence Action SummitAs the "AI revolution" begins to unfold within organizations, it's important to understand its impact now, monitor its development, and encourage the sharing of knowledge and best practices.In this context, the AI Action Summit aims to facilitate international exchanges on the future of work and build on existing initiatives in order to:improve our understanding of AI systems and their uses, to better anticipate their impact on the world of work, training and education;promote AI deployments to serve productivity skills development, and well-being at work.The ambition of the members mobilized on the "Future of Work" theme, as part of the Summit, is to promote the socially responsible use of AI through sustained social dialogue. Beyond the context of the Summit to share this information, work has been launched on an initiative to build capacity to measure the impact of AI and algorithmic management in the workplace.The aim is also to encourage economic and social players to commit to sustainable and responsible AI at the Summit, by bringing together nearly 40 government representatives from five continents, international social partners, academics and researchers, experts, economic players and companies.Exchanges at the Summit should therefore provide relevant tools for players wishing to embark on the path of responsible AI in the field of work, in particular by mobilizing tools that avoid gender discrimination in access to employment, as well as the deployment of a human resources policy that promotes training and adaptation to the new needs generated by AI.

• Innovation

Published on 11/02/25



[Feature] Why and how to make artificial intelligence frugal?

ForewordArtificial Intelligence (AI) systems consume a lot of energy. To operate, they need large computing capacities to train on billions of data sets, which requires powerful servers.The rise of artificial intelligence has led to such an increase in energy consumption that data centers now use more energy than 92% of the world's countries. Beyond the burden this increase places on power grids, the development of AI also threatens climate objectives.The quest for a more frugal, more sustainable or more durable AI is guiding a great deal of research and initiatives.Unlike traditional artificial intelligence models, which are often data- and compute-intensive, and therefore require a lot of electricity and servers, and contribute to warming the environment and the climate, frugal artificial intelligence favors lighter, more efficient approaches.Frugal AI is now among the four areas to be explored in the second phase of the national artificial intelligence strategy (SNIA), which covers the period 2022-2025, particularly in its research component. The French Ministry of Ecological Transition, for its part, is working to structure a Community of AI Actors in Territories, through a "Call for projects - Frugal AI demonstrators serving the ecological transition in territories". This ministry has also initiated a reference framework to measure and reduce the environmental impact of AI.Several voices have been raised in recent months to make frugal AI a priority, in France, of the national artificial intelligence strategy:In its assessment of the first phase of the SNIA, the Cour des Comptes (French National Audit Office ) sees frugal AI as a "founding principle that should permeate all the research funded by the second phase". However, the challenges of energy and data frugality are " little taken into account in research work", "beyond a few promising initiatives".According to the Economic, Social and Environmental Council (CESE) in an opinion issued in September 2024, " there is a concern that funding is lacking for AIs designed directly to protect the environment (...) Research efforts have only marginally focused on frugal AI". The CESE recommends focusing public funding on AI with a direct environmental purpose and on frugal AI.The Office parlementaire d'évaluation des choix scientifiques et technologiques (OPECST), for its part, makes the goal of frugal AI an "imperative".This theme of frugality is on the agenda of the Artificial Intelligence Summit to be held in Paris on February 11 and 12. A coalition for the environmental sustainability of AI could be formed at the end of the Summit. AI drives energy consumption through the roof"The conclusion of the Parliamentary Office for the Evaluation of Scientific and Technological Choices (OPECST), in a comprehensive report (339 pages), is that the situation is clear."AI requires considerable energy inputs throughout its lifecycle. AI tools, especially generative AI tools that use large language models (LLMs), require a lot of computing power to train their algorithms, and therefore electricity. The AI giants' supercomputers call on tens of thousands of Nvidia chips, even before they are used, just to train the models. While it is estimated that datacenters alone consume less than 2% of the world's electrical energy today, this figure could rise to 4% by the end of the decade under the impact of AI, whose installations require 3 to 5 times more electricity than traditional datacenters."Most models have to be trained on hundreds or even thousands of graphics processing units( GPUs). These GPUs enable fast matrix calculations, but with proportionately high energy consumption and CO2 production. For example, training GPT-3 on GPUs required 190,000 kWh, or 85,000 kg of CO2 equivalents, equivalent to driving a car 700,000 kilometers, or a round trip from the Earth to the Moon.According to the International Energy Agency (IEA), global electricity consumption in the AI and datacenter sectors could more than double by 2026 compared to 2022, reaching 1,050 terawatt-hours (TWh), equivalent to Japan's electricity consumption.This insatiable demand for energy could, according to Gartner, put energy suppliers under pressure and lead to shortages, limiting the growth of new data centers from 2026 onwards. Frugal AI finds its way into the national artificial intelligence strategyMathematician, and MP, Cédric Villani in his report on artificial intelligence (AI) questioned the ecological sustainability of AI back in 2018. "It's a question of thinking about a natively ecological AI and using it to better think about the impact of humans on their environment. This is a matter of urgency: by 2040, global storage space requirements, fundamentally correlated with the development of digital technology and AI, are likely to exceed global available silicon production (...) The greening of the AI value chain will necessarily involve open hardware and open software architectures which, in addition to being a confidence-building factor, can enable significant energy savings."The first phase (2018-2022) of the national artificial intelligence strategy (SNIA) aimed to equip France with competitive research capabilities. The first phase of the SNIA was funded to the tune of 1.5 billion euros. It notably enabled the creation and development of the network of interdisciplinary artificial intelligence institutes (3IA), the establishment of chairs of excellence and doctoral programs, as well as the Jean Zay supercomputer.In November 2021, a new "acceleration" phase of the SNIA was announced. Frugal AI is among the four priorities of this second phase, along with Embedded AI, Trusted AI, and Generative AI. This second phase, which covers the period 2022-2025 and is endowed with 2.2 billion euros, plans, in particular, to " support 10 frugal AI demonstrator or technological development projects".Frugal AI is also one of the priorities of the Artificial Intelligence Priority Research Program and Equipment (PEPR), co-piloted by CEA, CNRS and INRIA, with "the ambition of identifying software architectures, frugal learning methodologies for training on controlled databases, and advanced application optimization to rationalize energy consumption of machine learning-based AI systems as far as possible". " Frugal AI must permeate all research in a concrete way".In its first phase, the national AI strategy prioritized research. However, according to the Cour des Comptes, the challenges of energy and data frugality are " little taken into account in research work", "beyond a few promising initiatives". The use of algorithms seems to be the dimension most integrated into research work (see graph below). However, the Cour des Comptes sees frugal AI as a " founding principle a founding principle that should permeate all research funded by the second phase ". "Many efforts remain to be made in terms of frugal AI and trusted AI, both in terms of mobilizing researchers, allocating resources and the indispensable construction of a European normative framework," concludes the Cour des Comptes. "In order to ensure that frugal AI development objectives are met in the second phase, whatever the sector, it would be advisable to improve the measurement and monitoring of the ecological footprint of AI, and to build a framework of actions compatible with the control of the environmental footprint." "Concentrate funding on AI with a direct environmental purpose and on frugal AI".The French Economic, Social and Environmental Council (CESE), for its part, in an opinion issued in September 2024, observes that " AI is currently used mainly in areas such as finance, targeted marketing and now industry, and not to train and operate optimization algorithms that benefit the environment (...) Under these conditions, it is to be feared that funding will be lacking for AI designed directly to protect the environment".Six years after the Villani report, we have to admit that research efforts have only marginally focused on "frugal AI". While the Artificial Intelligence and Ecological Transition Roadmappublished in 2021 (by the French Ministry of Ecological Transition) includes a "frugality" section, there are still too few projects underway".According to the CESE, frugal AI could be a priority focus of the fourth Future Investment Program 2021-2026 and the France 2030 investment plan.The EESC recommends focusing public funding on AI with a direct environmental purpose and on frugal AI.A systematic assessment of the environmental footprint of AIs will have to be carried out, requiring the companies concerned to be transparent about resource consumption, and building an evaluation benchmark. Such an AI "Ecoscore" will enable users, citizens and companies alike, to choose the most sober AIs (recommendation no. 5).At the same time, engineers will have to become more involved in the eco-design of equipment, particularly terminals, the renewal of which is accelerated by AI (recommendation no. 6), and users can be made more aware of their responsibilities through information campaigns on the environmental footprint of AI and the possibility, currently non-existent, of disconnecting AI uses and data collection on their applications (recommendation no. 7).Last but not least, measures are immediately applicable to data centers. For data center projects, the public authorities must ensure compliance with the objective of zero net artificialisation, in particular by using sites that have already been artificialised (recommendation no. 8), the European sustainability rating system for these centers must be taken into account, and the principle of recovering waste heat must be integrated (recommendation no. 9). 12 frugal AI demonstrators in the regionsAs part of its "AI and ecological transition" roadmap, the French Ministry of Ecological Transition has been supporting frugal AI-based solutions for territories since 2022.The " Démonstrateurs d'AI frugale des territoires" (DIAT) call for projects targeted AI projects capable of making predictions and recommendations to aid decision-making. " AI solutions must be frugal in terms of data and energy, and their own environmental impact must be measured, which must be thought out upstream of project construction ".12 projects were selected for this call for projects. The Banque ds territoires, operator of the call for projects, has identified three approaches to frugality among the selected projects:Minimize data collection, aiming for quality rather than quantity, and ensuring that only the data needed for the purpose is updated". This was notably the case for the IA.rbre project, which uses AI to help the Lyon metropolitan area detect "plantable" areas and use vegetation to create cooling zones or combat runoff phenomena.Limiting the number of sensors: frugality can also be achieved by exploiting data that are not originally environmental data. Through the Predict AI'r project run by the Paris-Ouest La Défense public establishment, the consortium will exploit cell phone boundary data to deduce, using AI, users' travel patterns and their impact on air quality. A strategy that could make it possible to avoid multiplying the number of air sensors, while at the same time being applicable to any territory, in France and worldwide.Frugality can also be measured by the positive externalities that AI generates in organizations. In Saclay, for example, the Urba (IA) project supported by the agglomeration community promises to help local authorities in the Paris region revise their local urban planning schemes (PLU) by integrating all normative constraints in terms of environmental protection, whether regional or national.These calls for projects are enabling the Ministry of Ecological Transition to bring together a Community of AI Actors in Territories (CAIAT) made up of representatives of territories, competitiveness clusters, research institutions, VSEs/startups/SMEs and companies. A benchmark for measuring and reducing the environmental impact of AIThe Ecolab of the Commissariat Général au Développement Durable (CGDD) and the Association Française de Normalisation (AFNOR) have published a methodology for assessing the effect of an AI service on the environment. In addition to providing an initial operational reference framework for quantifying its impact, the document contains a set of best practices to adopt in moving towards frugal AI.The method has been developed using a global approach, i.e. it covers all the digital resources mobilized to deliver the service. Starting with data centers (servers, storage bays, network equipment), which host and process digital data. Next come the network infrastructures that ensure data transmission to users. And finally, the terminals (smartphones, computers, televisions, etc.) that operate, receive and consult content.The entire lifecycle of a system is taken into account, from design to development to commissioning. This implies a reflection on the very necessity of using AI versus less resource-intensive solutions. AI must not become a systematic response to every need.Frugality also means minimizing the material and energy resources mobilized to create the AI system. This includes rationalizing AI models, reducing the volume of data used to train them, and optimizing the infrastructures they mobilize, such as data centers. The compression of algorithms and the use of open source data are also encouraged to reduce their carbon footprint.Throughout the AI lifecycle, the standard then requires measurement of the direct impact of the AI (energy consumed, CO2 emitted, resources consumed) as well as its indirect effects, such as the rebound effect, where an improvement in efficiency can lead to an increase in consumption.Finally, clear governance must be put in place to ensure that frugality principles are integrated into every stage of an AI project. AI suppliers and producers must communicate transparently on the actions implemented to limit impacts, providing precise assessments and clear methodologies.

• Data

• Innovation

• Environment

Published on 07/01/25