Executive summary

Executive summary Part 1 — of public sector information must Building a Data-Focused provide an opportunity to speed up the opening of public data and outline Economic Policy the terms and conditions for access to personal data on public interest grounds. The current reform of EU In this area AI heavyweights, such as copyright rules should at last China and the US, and emerging AI authorize text and data mining and powers, such as the UK Canada and enable our public research to be more Israel, are developing extremely competitive. different approaches. Thus, France This data policy must be designed and Europe will not necessarily take with the aim of safeguarding their place on the world AI stage by sovereignty: it is vital for France and creating a “European Google”, Europe to maintain a firm stance on instead they must design their own data transfer outside the European tailored model. Union. The AI strategy must also European Data Ecosystem capitalize on the high protection A whole range of uses and standards enshrined in the incoming applications rely on the availability of European General Data Protection data, so this is usually the starting Regulation (GDPR). Recent laws on 2 point for any AI-based strategy. Yet individuals’ rights to data portability data currently mostly benefit just a could therefore be part of a broader handful of very large operators, so citizen-based rationale, to enable the greater data access and circulation State and local authorities to recover will be required to restore a more data with the aim of developing AI- even balance of power by extending based uses for public policy purposes. these benefits to government Raising Visibility for AI Players authorities, as well as smaller France has all the required assets to economic actors and public research. take its rightful place on the For this to happen, the public international arena, yet our authorities must introduce new ways companies and academic networks of producing, sharing and governing suffer from a lack of visibility both in data by making data a common Europe and overseas. Large 1 good . This should involve companies sometimes opt to rely on encouraging economic players to dominant world actors in the sector, share and pool their data, with the rather than entrusting their data to our State acting as a trusted third party. In home-grown talent, either because some circumstances, public they are not aware of this wealth of authorities could impose openness on skills within the country or because certain data of public interest. they prefer to adopt a very cautious Meanwhile in Europe, a number of approach. Our mission therefore reforms currently underway must suggests bringing together French AI provide for greater access and wider actors under a unique and strong circulation of data. The forthcoming banner, which would include revision to the directive on the re-use certifications and “innovation in the 1. Common goods refer to resources where 2. Users’ ability to receive their personal use and governance are defined by a data for their own use or to transmit to community. another data controller. 2

The Report in 10 Pages field” awards aimed at singling out emission-free urban transport, etc. the most innovative AI solutions and These various business policy issues attracting potential buyers. and challenges, each specific to its This approach must also be set own sector, go beyond the alongside a more organized approach boundaries of AI, but could help to demand for AI, which could involve provide a ripe breeding ground for its the creation of an information one- development. stop shop aimed at helping potential The second key point of this strategy AI buyers outline their requirements involves setting up shared sector more effectively and ascertain the platforms, which must provide secure companies that could best address and tailored access for the various their needs. participants in these different A Clear Policy to Focus on Four ecosystems (researchers, companies, Strategic Sectors public authorities) to useful data for the development of AI, as well as to It is vital to take advantage of our software resources and extensive economy’s comparative advantages computing infrastructure. In a public- and its areas of excellence in order to private continuum, these platforms bolster the French and European must enable the various stakeholders artificial intelligence ecosystem. In this to develop new functionalities that are respect, our task force recommends tailored to the individual features of avoiding spreading efforts too thinly, each sector. but rather focus on four key sectors: Lastly, it is vital to streamline the AI healthcare, environment, transport- innovation track with the mobility and defense-security. These implementation of innovation sectors are all crucial from a public sandboxes, involving three key interest standpoint, all require strong features: a temporary easing in certain impetus from the State, and they can regulatory restrictions in order to give all be the focus of interest and free rein to innovation, support for ongoing involvement from public and participants as they address their private stakeholders. obligations and lastly resources for The business strategy for each of use in field testing. these sectors must allow for the The State Both Transforms and creation and organization of Shows the Way ecosystems based on the different major sectoral challenges. Artificial It is vital for the State to be a key driver intelligence should not be developed in these various areas of as an objective or an end in itself, but transformation. Public authorities rather it must be a way to channel this must ensure that they adopt the energy to develop practical necessary material and human applications and uses that help resources to factor AI into the way improve our economic performances they address public policy, with the while contributing to the public aim of both pursuing modernization interest i.e. early detection of and acting as an example to be 3 diseases, the 4 Ps of healthcare , followed. elimination of medical deserts, 3. Personalized, preventive, predictive and participatory healthcare.

Executive summary This transformation will obviously take up a network of independent but time and the various ministries and coordinating Interdisciplinary government bodies display varying Institutes for Artificial Intelligence degrees of progress in the field of AI. within defined number of public An inter-ministerial coordinator role higher education institutions. These should therefore be created, devoted bodies would house researchers, to implementing this strategy, with engineers and students, and should support from a shared specialist be located all across the country, each center consisting of around thirty staff one devoted to specific aspects of AI, tasked with acting in an advisory and with a very strong focus on an capacity for the different government interdisciplinary approach, notably by bodies. including social scientists. Meanwhile, public procurement First and foremost, it will be crucial to needs to be reviewed: this budget is attract French and international estimated at close to 70 billion euros academics, and these institutes will for the State, public authorities and therefore have to create an attractive local bodies each year and it is working environment in order to insufficiently oriented towards effectively address competition from innovation. Our task force “Big Tech”. They should therefore be recommends a number of measures set up as AI “free zones”, with a aimed at using public procurement to considerable reduction in support European industries and at administrative formalities across the breathing fresh momentum into board, hefty salary top-ups, and innovative public spending. support in improving quality of living. These institutes could offer full-time positions as well as intermediary Part 2 — affiliate status for researchers who Promoting Agile and remain in founding establishments. Enabling Research It will also be important to attract private partners, such as large groups, SMEs and start-ups, which can deliver The French academic research is at brand new AI solutions, by enabling the forefront of worldwide exploration them to train their own engineers, on mathematics and artificial recruit premium quality engineers, intelligence, but the country’s and make or consolidate scientific progress does not always technological breakthroughs. A range translate into concrete industrial and of options could be provided to economic applications. The country is enable participants to get involved on hit by the brain drain towards US a tailored basis, based on heavyweights, and training personalized framework contracts that capabilities on AI and data science fall provide for a simple fast-track well short of requirements. cooperation process. Bringing Academics Together These institutes should heavily invest Within Interdisciplinary Research to increase the supply of attractive Institutes on Artificial Intelligence and diversified AI training programmes. The presence of It is key to bolster our position internationally renowned academics worldwide on AI research by setting with the support of premium teams,

The Report in 10 Pages the opportunity to interact with world- Make Public Research Careers More class corporations via internships and Attractive innovation competitions, multi- It is unrealistic to try to compete with disciplinary training programmes with GAFAM’s salary scale, but the gap is joint degrees, and scholarships for currently so wide that it tends to Masters’ degree and Ph.D. students discourage young graduates, even should help significantly boost the those who are extremely interested in number of students taking AI training public research and contributing to at these institutes. the common good to join public Lastly, it is essential to take a nation- research institutions. Doubling wide approach to coordinate this salaries in the early stages of their interdisciplinary institute network careers at the very least is a vital from both scientific and administrative starting point, otherwise the pool of standpoints, in order to ensure that young graduates interested in higher they are run efficiently and education and academic research will transparently. From a scientific definitively dry up. It is also important standpoint, this involves the to make France more attractive to coordination of seminars, pooling expatriate or foreign talents, with training resources, coordination of financial incentives for example. internships and consolidation of their results. Meanwhile, in administrative terms, this will involve assessing the Part 3 — red-tape fast-track provisions granted Assessing the Effects of to all institutes and ensuring that each one benefits from this set-up, while AI on the Future of Work keeping procedures streamlined and and the Labor Market, ensuring that each institute can operate independently. and Experiment Research Computing Resources Adequate Policy AI research institutes need to have the Responses computing resources required to compete with the virtually unlimited resources of private dominant actors. The labor market is undergoing vast To do so, our task force therefore changes, but it is not yet fully suggests setting up a supercomputer equipped to address it. There are designed specifically for AI usage and considerable uncertainties on the devoted to researchers and their effects of the development of artificial economic partners during their shared intelligence, automation and robotics, initiatives. particularly on job creation and This supercomputer is vital but should destruction. However, it looks also be rounded out by an access increasingly certain that most sectors package to a private cloud set-up, and companies will be widely developed European-wide and reshaped. We are entering a new era tailored to meet the specific features of major technological transition and of AI in terms of computing time and history shows us that previous periods data storage space. of transition did not always run smoothly. Indeed, they sometimes involved drastic political

Executive summary readjustment, which often hit the complex and difficult to steer. For most fragile portions of the example, professional training is population the hardest. So it is worth 32 billion euros per year, with a important to face this issue head-on vast array of funding channels and a and take resolute action, while not whole range of different stakeholders giving in to panic or fatalism. involved. This firstly involves looking into the It is therefore crucial to create a space complementarity between humans where both prospective capacities, and artificial intelligence: if we are to macroeconomic forecasts and assume that, for most jobs, individuals analysis of changes in uses can be will have to work with a machine, then linked to concrete experimentation it is vital to find a complementarity capacities articulated with actions set-up that does not alienate staff but aimed at certain categories of instead allows for the development of workers. A permanent structure could truly human capabilities, such as therefore be created to spearhead creativity, manual dexterity, problem- these subjects within labor and solving abilities, etc. This can take professional training public policy, several forms. Firstly, it might involve with a twofold role: to anticipate and a shift in labor relations to fully experiment. integrate digital challenges and This experimental approach can then develop a ‘positive complementarity be used to initiate logics different index’. More broadly speaking, from those currently in force in legislation could be implemented to vocational training, i.e. it is now deal with working conditions at a time broadly left up to employees, who of increasing automation in order to take personal responsibility for their factor in new risks. Lastly, formal own training. Yet in light of the education and lifelong learning potentially swift or even exponential should be overhauled in order to speed of transformation, it is difficult promote experimental teaching for current general programmes to methods that can help graduates and incorporate all possible situations and staff develop the creative skills that take on board both the requirements are becoming increasingly vital. of the entire population and the need Setting up a Public Lab for Labor for a fast but targeted approach. Transformations Furthermore, staff do not all react in The top priority is to ensure that the the same way to the transformation of ability to anticipate is sustainable, their jobs and do not all have the continuous and above all articulated same ability to build a new career with public policies. The publication path. of studies on the future of the labor In this respect, trials could be carried market often sparks off fascinating out to design programmes that target collective debate, but does not always certain groups, whose jobs are result in concrete actions, with public deemed to be more at risk from policy being only slightly adapted automation and who would have without fully taking into account the more difficulty addressing their results of these forecasting exercises professional development without yet. Transformation can be extremely guidance. This approach involves fast, while public policy moving somewhat away from the implementation procedures are current strategy whereby employees

The Report in 10 Pages alone are responsible for their own new courses on AI on the other e.g. career development. law-AI joint degrees, general Trying out New Professional modules, etc. All degree courses Training Funding Methods to should be involved, i.e. 2-year, 3-year, Successfully Deal with Value Masters, Ph. D, etc. Transfer Funding for staff training is calculated Part 4 — on the basis of a company’s total Artificial intelligence payroll, yet the development of AI further promotes the transformation Working for a More in value chains and reduce the link Ecological Economy between those funding professional training and those who derive the value-added from it. Companies with Carving out a meaningful role for a very small payroll can therefore artificial intelligence also means create a large portion of the value- addressing its sustainability, added in an overall value chain that especially from an ecological they are responsible for extensively standpoint. This does not just mean changing, e.g. by developing considering the application of AI in software for self-driving cars. Yet for our ecological transition, but rather the moment, they do not take part in designing natively ecological AI and funding the career transition of staff using it to tackle the impact of human employed by other companies that action on the environment. This is an operate across the value chain. urgent matter as world data storage We therefore propose initiating requirements, inherently correlated to dialogue with industrial partners on the development of digital how value-added is shared across the technology and AI, could exceed entire value chain. This type of available worldwide silicon negotiation cannot be based on the production out to 2040. usual formats for social dialogue, First and foremost, France and Europe which mostly operate nationwide with can spearhead this smart ecological a vocational branch approach. Trials transition by raising awareness on the could be organized by the international arena. The primary task International Labor Organization or is to consider both the impact of AI on sector social dialogue committees achievement of the UN’s sustainable focused on products and value chains development goals, how it puts that are particularly affected by these pressure on certain goals and how it value questions. can accelerate others. AI must be Training Talents in AI at Each and included in initiatives emerging as Every Degree Level part of the Paris Climate agreement One clear target must be set: triple and the Global Pact for the the number of people trained in Environment. artificial intelligence in France in the Players in both digital and ecological next three years, by ensuring that transition must join forces, which existing training programmes focus require setting up a devoted space for more on AI on the one hand, but also AI research and energy resource by setting up new programmes and optimization research to meet, and

Executive summary promoting projects at the crossroads Dissemination of Ecological Data of life sciences and ecology, climate The development of green AI is only and weather research. feasible if ecological data can be Consumers must also play a part in open. So it is vital to make currently making these technologies greener. available public data open to all, both Our task force therefore proposes the researchers and European companies creation of a platform devoted to alike, out to 2019 in order to develop assessing the environmental impact of AI solutions to promote ecological smart digital solutions. This platform transition i.e. data on weather, should also include a simple calculator agriculture, transport, energy, to enable all citizens to gain greater biodiversity, climate, waste, land awareness of these impacts and registry and energy performance compare the environmental footprint assessments. Access to more sensitive of the various products, services, data could be managed on the basis software and hardware. of more specific situations, e.g. to Fostering Greener AI address sector challenges. It is also important to open privately-owned It is also important to tackle data where necessary. breakthrough innovation in the semiconductor sector, one of the physical building blocks of AI. In this Part 5 — respect, neuromorphic4 technology Ethical Considerations can allow for considerable energy savings, and France is already a of AI pioneer in this area. Public authorities must also act to make the value chain greener and Recent AI-led progress across a support the European cloud industry number of sectors (self-driving cars, to promote its ecological transition. image recognition, virtual assistants) Some market participants already and its increasing influence on our provide excellent examples of energy lives are driving public debate on the optimization and these best practices issue. This debate included extensive now need to be extended to the analysis of the ethical challenges entire sector. A certification process raised by the development of artificial could also be set up to reward the intelligence technologies and more most outstanding solutions. broadly speaking by algorithms. Far Lastly, making the AI value chain from the speculative considerations greener will clearly require open on the existential threats of AI for hardware and open software, which humanity, the debate seems to focus are not only a confidence indicator on algorithms that are already present but can also lead to significant energy in our daily lives and that can have a savings and provide inspiration for major impact on our day-to-day initiatives currently underway in existence. Europe. If we want to develop AI technologies that comply with our values and social 4. Neuromorphic chips are based on the workings of the human brain.

The Report in 10 Pages norms, then it is vital to act now to legal proceedings, during an rally round the scientific community, investigation undertaken by an public authorities, industry, business independent administrative authority owners and civil society organizations. or on request by the Defender of Our mission has endeavored to put Rights (Défenseur des Droits). forward some humble suggestions Implementing Ethics by Design that could lay the foundations for the ethical development of AI and Research staff, engineers and promote debate on this issue within business owners who contribute to society at large. designing, developing and marketing Opening the Black Box AI systems play a decisive role in tomorrow’s digital society, so it is vital A large proportion of ethical that they act responsibly and factor in considerations are raised by the lack the socio-economic effects of their of transparency of these technologies. actions. With this in mind, it is AI provides spectacular results for important to make them aware of the reasons that researchers sometimes ethical issues involved in the have difficulty to explain: this is known development of digital technologies as the black box phenomenon, where right from the start of their training. we can see input data and output data This aspect is lacking in today’s for algorithm-based systems, but we courses at engineering school and in do not really understand what exactly universities’ IT programmes, yet the happens in between. AI can extent and complexity of ethical reproduce bias and discrimination issues these future graduates will face and is becoming increasingly present continue to grow. in our social and economic Looking beyond engineer training, environments, so opening the black ethical considerations must be fully box is a key democratic issue. factored into the development of Explaining machine-learning artificial intelligence algorithms. A algorithms has become a very urgent discrimination impact assessment matter and is now actually a separate could be introduced, similar to the field of research, which must be privacy impact assessments already supported by public authorities. made compulsory by General Data Three areas in particular require an Protection Regulation for some data extra focus: obviously the production processing. The overarching aim here of more explicable models, but also is very simple: have AI developers the production of more intelligible consider the right questions at the user interfaces and an understanding right time. of the cognitive mechanisms used to More broadly speaking, the produce a satisfactory explanation. increasing use of AI in some sensitive Transparency is clearly key, but areas such as policing, banking, looking beyond this issue, it is also insurance, the courts and in Defense vital to facilitate audits of AI systems. (with the question of autonomous This could involve the creation of a weapons) raises a real society-wide group of certified public experts who debate and implies an analysis of the can conduct audits of algorithms and issue of human responsibility. We databases and carry out testing using must also consider the role of any methods required. These experts automation in human decisions: are could be called on in the event of there areas where human judgement,

Executive summary fallible though it is, must not be Parity and Diversity: Acting to replaced by a machine? Promote Equality Setting Up an AI Ethics Committee Despite the slow but steady Our mission recommends the creation feminization of scientific and technical of a digital technology and AI ethics sectors, digital technologies remain committee that is open to society. something of an exception, with This body would be in charge of gender balance still very far off. As leading public discussion in a digital technologies and, in the very transparent way, and organized and near future, artificial intelligence governed by law. It should work become widely present in our lives, alongside sector committees and this lack of diversity can lead combine short-term considerations, algorithms to reproduce often such as economic and industrial unconscious cognitive bias in impacts, with the ability to take a step programme design, data analysis and back and take the long view. the interpretation of results. One of the major challenges of AI is ensuring Recommendations from the greater representation within our committee, which would operate societies. entirely independently, could help Educational efforts on equality and inform researchers’, economic digital technology are obviously vital, players’, industry’s and the State’s but greater diversity could also be technological decisions. Its achieved with an incentive policy recommendations could act as a aimed at achieving 40% of female benchmark for resolving ethical students in digital subject areas in matters (e.g. on self-driving vehicles) universities, business schools and and hence provide a standard for AI their preparatory classes out to 2020. developments. All moves to promote diversity in digital companies could be further Part 6 — fostered by a nation-wide approach to promote diversity in technology via a Inclusive and Diverse AI national database aimed at documenting gender inequality in the workplace and the provision of funds Artificial intelligence must not devoted to supporting diversity in AI. become a new way of excluding parts Developing Digital Mediation and of the population. At a time when Social Innovation to Ensure AI these technologies are becoming the Benefits All keys to opening the world of the future, this is a democratic Given the extent of future AI-led requirement. AI creates vast transformation, we have a collective opportunities for value creation and responsibility to ensure that no-one the development of our societies and gets left behind. For everyone to truly individuals, but these opportunities benefit from breakthroughs made in must benefit everyone across the AI, our procedures for access to rights board. must change and our mediation capabilities must also be considerably bolstered. So our mission puts forward a proposal to set up an

The Report in 10 Pages automated system to help manage remain very focused within a small administrative formalities, aimed at number of companies. Setting aside improving public awareness of healthcare, social fields receive only a administrative regulations and how tiny portion of private investment. they apply to each individual’s This set-up for the AI-led innovation personal situation. In addition, fresh ecosystem has consequences on the mediation capabilities must be speed of progress made in social developed to support those who matters. In order to redistribute these require help, in cooperation with care innovation capabilities, public networks already present nation-wide. authorities could embark on specific Lastly, it is crucial that public programmes to support AI innovation authorities support the development in the social arena and provide the of AI-based initiatives in the social necessary systems for the various arena. AI-led innovation capabilities parties in the sector so that they can benefit from AI-related progress.