Technological results
The Body of Knowledge (BoK), the cornerstone of trustworthy AI
In a world where Artificial Intelligence is emerging as a transformative force, the Body of Knowledge is an essential body of knowledge. This unique repository, resulting from the work of the Confiance.ai program, gathers and organizes essential knowledge on the principles and practices of trustworthy AI.
The BoK covers the end-to-end development process of a system based on trustworthy AI. More precisely, it takes the case of a system whose functionality we wish to automate through the involvement of Machine Learning technology, while guaranteeing a good level of trust in this functionality.
In particular, it addresses crucial challenges such as the engineering of data serving as the basis for Machine Learning, the robustness of ML models, their explicability, their embeddability, the quantification of their uncertainty, and their monitoring. But also the definition of the operational conditions under which AI-based system functionality is designed to function: the ODD, and the definition of a global V&V plan based on Assurance Cases. These topics are essential for building AI systems worthy of trust.
- The Body of Knowledge website responds to these challenges by:
Centralizing trustworthy AI knowledge; - Offering resources to support engineering (methods, tools);
- Serving as a reference for professionals, researchers and students;
- Promoting the work of the Confiance.AI program for the development of trustworthy AI.
The BoK offers invaluable access to reliable, up-to-date resources. It is the reference guide for developing trustworthy AI, anticipating implications and risks.
The BoK demonstrates that trustworthy AI is not a brake on innovation, but a catalyst for sustainable progress. By bringing together knowledge on this topic, it lays the foundation for the future of trustworthy AI.
Explore the BoK now and join the movement for AI that is responsible, trustworthy and beneficial to all.
Resource catalog
Intended to provide information to anyone considering integrating trusted artificial intelligence into their systems, the Confiance.ai catalog compiles a set of resources produced or evaluated as part of the program. Based on a philosophy of openness and free consultation, the catalog offers resources based essentially on open-source licenses; these resources, whether documentary or software, can be used and integrated into trusted artificial intelligence design processes.
By browsing the resource catalog you will find:
- Methodologies that describe the essential steps for integrating trusted AI into existing engineering processes;
- Tools designed to facilitate the application of these methodologies by engineers at different stages;
- Evaluation reports helping to estimate the performance of tools in different application contexts;
Comprehensive reviews covering the different topics of trusted AI engineering.
If these resources are naturally classified by typology, the catalog also offers the possibility of browsing them according to the role, the theme, the trust attribute or even the stage of the AI engineering process that interests you.
