Description du projet
Une initiative pour exploiter les incertitudes de l’IA dans la région du Grand Paris
Dans le paysage en évolution rapide de l’IA, le défi de l’évaluation de la fiabilité des décisions se pose avec acuité, touchant des domaines allant des simulations numériques aux réactions des systèmes dans le monde réel. Les incertitudes liées aux données d’entrée et aux résultats des modèles créent un besoin pressant de confiance dans les applications de l’IA. Avec le soutien du programme Actions Marie Skłodowska-Curie, le projet DeMythif.AI vise à gérer les incertitudes, à assurer l’explicabilité et à promouvoir la frugalité. Le projet est à l’origine d’une approche transformatrice de l’IA, permettant à 30 doctorants du Grand Paris d’aborder et de surmonter les obstacles inhérents à l’avancement de la technologie. La force du projet réside dans la diversité des approches interdisciplinaires, la promotion de collaborations de recherche dynamiques et la promesse d’un avenir où les applications de l’IA inspirent confiance et compréhension.
Objectif
Artificial intelligence is a pervasive and ubiquitous technology, with fast developments and ever-growing applicative horizons. While achieving outstanding results, estimating the decisions’ reliability is a critical and open problem with strategic outcomes: from estimating uncertainty in numerical simulations to harnessing system reactions in open environments. Uncertainties both in input data and model output should be handled to increase the confidence in AI applications : physics-based models, edge computing, data frugal approaches, interactions with humans, with many applications in energy, climate change adaptation, bioinformatics, engineering, fundamental and material science,… DeMythif.IA is an international doctoral training and career development program driven by UPSaclay, and its 19 research and industrial partners supported by thematic networks to enhance the scientific excellence and career development of 30 PhD fellows in Greater Paris area. The program focuses on 3 interdisciplinary scientific axes 1) handling uncertainties on data and model and quantifying uncertainties on the predictions, 2) managing explainability, so that the trained model outcome can be trusted and explained to a human, and 3) encouraging frugality, both in term of labelled data and energy for training, so that real-life applications (as opposed to proofs-of-concept) emerge. The fellows will benefit from world-class scientific programs, quality supervision and industrial secondments and also from personalized career development activities and transferable skills training (interpersonal, communication, digital, entrepreneurship, open science, gender, ethics). DeMythif.AI’s strength lies in its diversity permitting to fine-tune the content offered to the fellows from a professional perspective. It ultimately allows concretizing dynamic research collaborations and developing strong synergies around the academic and industrial community, opening new perspectives beyond the project.
Champ scientifique
Mots‑clés
Programme(s)
- HORIZON.1.2 - Marie Skłodowska-Curie Actions (MSCA) Main Programme
Régime de financement
HORIZON-TMA-MSCA-Cofund-D - HORIZON TMA MSCA Cofund Doctoral programmeCoordinateur
91190 Gif-Sur-Yvette
France