Launched in January 2021, the Confiance.ai program is a technological pillar of the Grand Challenge “to secure, certify and make reliable systems based on artificial intelligence”. On the occasion of the first Annual Day held in Toulouse on 6 October 2021, initial results were unveiled, showing a sustained dynamic and an increasingly rich ecosystem of partners.

Led by the SystemX Institute of Technological Research (IRT), the Confiance.ai program is driven by a group of 13 French industrialists and academics (Air Liquide, Airbus, Atos, Naval Group, Renault, Safran, Sopra Steria, Thales, Valeo, as well as the CEA, Inria, the IRT Saint Exupéry and the IRT SystemX) and has been allocated 45 million euros over 4 years. It aims to meet the challenge of industrializing artificial intelligence (AI) in critical products and services.

“Trusted AI is one of the key current issues in the industrial world, as AI is spreading so rapidly in all sectors and especially those that do not tolerate decision errors. We can only be extremely satisfied with the momentum that has been built up in France as part of the Confiance.ai program.

Fully operational since January, it has attracted a large number of SMEs and academic players who have joined our Common House and contributed their expertise and technologies, in collaboration with the 13 structuring partners who initiated the programme”, comments Julien Chiaroni, Director of the Grand Défi IA de confiance at the General Secretariat for Investment.

A first version of the trust environment by the end of the year integrating more than 20 technological or methodological components

To meet these challenges, the program’s partners have focused on six initial use cases: understanding scenes from a camera, vision-based inspection of welds, prediction of liquid air demand, interpretation of aerial photos, visual industrial control, and airborne collision avoidance for unmanned aircraft systems.

These practical use cases have made it possible to assess the relevance of the first 20 technological components or methodological building blocks (e.g. evaluation of neural network robustness, quantification of prediction confidence intervals, generation of models robust to certain disturbances, methods for constructing and characterizing data sets, etc.), which have already been integrated into the first version of the confidence environment developed and which will be made available to partners by the end of the year.

This environment will eventually offer a sovereign, open, interoperable and sustainable software tool platform for the design, validation, qualification, deployment and maintenance of critical AI-based products and services.

An ecosystem of over 40 partners by the end of the year

Confiance.ai is part of an open and integrative approach. The program is now developing and federating the national ecosystem around trusted AI and has opened up to many associated partners.

Beyond the 13 structuring actors of the program, 11 laboratories have recently joined the collective after being selected in the framework of an AMI (call for expression of interest) focused on scientific challenges: Institut de mathématiques de Toulouse – ANITI, CRIL – Université d’Artois, CRISTAL – CNRS, LAMIH/UPHF / CRISTAL – Université de Lille, Inria KAIROS, IRIT – Université de Toulouse, LIP6 – Sorbonne Université, LITEM – Medial Lab – IMT, LITIS – Insa Rouen, ONERA and U2IS – ENSTA Paris. They will contribute to the maturation of scientific work or to the resolution of upstream scientific obstacles, most often in the form of a thesis or doctorate. To date, the program has already included 9 theses and 4 post-docs.

Other partners such as Apsys, LNE, Numalis and ONERA have also joined the programme, contributing their expertise and technologies.

Another AMI for deeptech start-ups and innovative SMEs was launched last July, and should enable a dozen French nuggets to be included in the programme by the 4th quarter of 2021.

In total, a powerful collective of more than 40 partners will collaborate on the sites of Paris Saclay and Toulouse, two important places of trust AI, around the development of this trust environment where everyone will be able to get value from this work. Confiance.ai is fortunate to be able to rely on this ecosystem and in particular on the Franco-Quebec program DEEL, the 3IA Aniti and the DataIA initiative.

A program that contributes to the European Commission’s “AI Act” regulation project

The European Commission has proposed a legal framework to promote trusted AI. However, its operational implementation requires the development of technical solutions allowing manufacturers to respond. Confiance.ai will allow the creation of a technical environment (on certification, reliability, evaluation, transparency and auditability of algorithms and systems) that will guarantee, tomorrow, a high level of trust in AI technologies.

Translated from Le programme Confiance.ai dévoile de premiers résultats significatifs avec 6 cas d’usage pratiques