This event served as an opportunity to engage with experts from various fields (refer to the end of the article) on the challenges and opportunities that AI presents for the industry. We discussed the importance of building trustworthy AI systems that uphold our values in terms of ethics, privacy, and transparency. Topics such as data governance, bias, interpretability, monitoring, measuring carbon footprint, as well as the need for vigilance and acculturation on these subjects were addressed to enhance our control over our AI algorithms and thus, increase our users’ trust.
This exchange was particularly meaningful as a follow-up to our work initiated in 2018 and culminated in 2023 with the accreditation of our project methodology (LNE) and an advanced level “Trustworthy and Responsible AI” certification (LabelIA Labs). These voluntary initiatives allowed us to highlight our methodology for building AI models capable of meeting the practical needs of various industries in production, while adhering to stringent standards. Numerous tools were implemented, including the formalization of AI’s ethical guidelines from the project’s outset, systematic bias measurement studies, methodical model comparisons to find a balance between performance and complexity, monitoring methods for a model in production to control its performance at all times, as well as the measurement and limitation of environmental impact.
At the DataLab Groupe, we continue to work on these topics to shape the future of trustworthy and responsible industrial AI.
Participants:
- Emeric Tonnelier: Tech Lead Analytics AI Squad, Crédit Agricole DataLab Groupe
- Matthieu Capron: AI Design Authority Responsible, Crédit Agricole Group
- Richard Eudes: Managing Director – Data, Advanced Analytics & AI, Deloitte
- Emilie Sirvent-Hien: Responsible AI Program Manager, Orange
- Nicolas Marescaux: Deputy Director, Member Needs Response & Innovation, MACIF Insurance
- Roxana Rugina: Executive Director, Impact AI