Numeum launches “Ethical AI”: the first initiative to operationalize ethical principles in Artificial Intelligence and promote trustworthy solutions

The initiative rests on three pillars: a manifesto to formalize the commitment of players, a practical guide to draw on a methodology and tools, and a community to move forward together. Supported by the Coordinator of the National Strategy for AI, Renaud VEDEL, the initiative by Numeum and its partners federates a network of players from different AI universes.

Ethical AI: a unique initiative to meet the new challenges posed by AI

Artificial Intelligence (AI) holds a great deal of promise, helping to meet major societal and environmental challenges, optimize business operations, diagnose faster and better, and more. However, certain technologies – particularly the most recent deep learning technologies – can give rise to mistrust and legitimate public questioning. It is essential to think differently about AI solutions, in order to meet expectations in terms of trust solutions and commitment to responsible digital.

Ethical AI is a unique project that brings together a network of partners (professional organizations, schools, interdisciplinary AI institutes, associations, foundations, think tanks…) committed to a more responsible digital world. Based on one of the recommendations of the Villani report, its ambition is to support “AI makers” who wish to integrate ethics at the heart of their activities, to support the operational translation of general ethical principles, derived from existing work, into practical methods and tools. The added value of more responsible solutions thus created will contribute to the strategic differentiation of technology developed in France on the international scene, and make it more competitive.

It is now necessary to improve practices at all levels (training, design, development, deployment…) to act more efficiently and in a spirit of continuous improvement.

A practical guide that is as close as possible to the real issues at stake, thanks to the co-construction work carried out with partners

The many exchanges and experience-sharing sessions resulted in a Practical Guide that proposes auseful method for operationalizing major ethical principles, during the design, development and deployment of AI solutions. 6 collective intelligence sessions were held at the end of 2020, bringing together 360 participants (coders, developers, data scientists, data architects, data engineers, AI product managers, testers, DPOs, lawyers, etc.). They all took part in brainstorming workshops to dissect the main theoretical principles of ethical AI and imagine ways of applying them in practice to the various components of the technology (algorithms, data, etc.). Made by and for those working on AI projects, the Practical Guide looks back at 6 key themes: loyalty, respect, transparency, fairness, control and safety.

A manifesto to formalize the commitment of players in the French digital ecosystem

En s’engageant autour dumanifeste Ethical AI, les signataires ont accès gratuitement auGuide pratique pour des IA éthiqueset intègrent la communauté des acteurs engagés pour le développement et la promotion de solutions d’IA plus responsables. Entreprises, organisations de l’écosystème du numérique ou acteurs de la formation aux systèmes d’IA, tous peuvent formaliser leur engagement en signant le Manifeste.

Les signataires du manifeste pour des IA éthiques s’engagent à :

• FAIRE: Mettre en œuvrela méthodologie et, dans la mesure de leur applicabilité,lesrecommandations présentées dans leGuide pratique pour des IA éthiquesédité par Numeum et coconstruit avec les représentants de l’écosystème de l’IA en France.
• COMMUNIQUER: Valoriser et promouvoir la dimension éthique des solutions d’IAque la structure conçoit et/ou déploie et/ou administre et/ou auxquelles elle forme.
• PROGRESSER: Contribuer à l’amélioration de l’état de l’art en matière d’éthique en IA, notamment par la participation d’un représentant de l’organisation aux travaux de révision et d’enrichissement du guide. Ces travaux s’appuieront sur les éléments partagés par les membres du collectif des signataires du Manifeste – ex. : retours d’expérience, avancées théoriques, etc.

Nous ont déjà rejoints: Accenture, Aivancity, AIvidence, Alter Way, Astek, CREEIA, Datavaloris, Devoteam, Docaposte, EITA Consulting, Horizon Software, Keyrus, La Clinique De L’Ethique (Factonics), Linagora, Linedata, Myriad, Orange Business Services, Quantmetry, Saaswedo, Sage, Seenapsys, TAAS, Talkr.ai by Kwalys, The Neoshields, The Voice Lab, Utocat, Ysance.

Pour en savoir plus, rendez-vous sur www.ai-ethical.com

Vous pouvez télécharger le Guide pratique ici et accéder au manifeste ici.