Scientific contributions to the Confiance.ai program
Scientific focus in Confiance.ai
Confiance.ai is an industrial research and development programme, its purpose was not to develop new science. Its main focus was to exploit available scientific and technological results, assess them and possibly extend them on industrially significant use cases demonstrating their benefits. Naturally, a number of software tools – individual components and large platforms – were developed and extended during the programme. These results can be consulted in Confiance.ai’s catalog, also available online.
Another result of the programme is the methodology for engineering AI-based trustworthy applications, directed towards machine learning systems. This methodology can be consulted in Confiance.ai Body of Knowledge website.
From its inception it was clear that the programme and its participants would benefit from a constant focus on scientific aspects of modern AI, both looking at inside developments and at the huge amount of science taking place outside. Even before Confiance.ai formally started, the major scientific challenges of the programme were identified. A scientific challenge is a subject on which there is no complete solution (methods, algorithms, ready-to-use tools) and which therefore requires additional research. There may be partial solutions, for example described in scientific publications, or even implemented within software tools, but these solutions cannot be used as they are and only resolve part of the subject
More than 75 individual challenges were identified, grouped into three large categories:
- trust and systems engineering with AI components;
- trust and learning data;
- trust and human interaction.
Publications and the Confiance.ai HAL collection
The HAL collection of Confiance.ai contains around 70 publications. Two of these publications are collections of papers and posters presented at the annual Confiance.ai days. Two white papers – available on the Confiance.ai website – presenting the scientific and technical challenges of the programme and the solutions developed.
The Confiance.ai program has focused on exploiting existing scientific and technological results to develop tools and methodologies dedicated to trusted AI. With over 75 scientific challenges identified and 70 publications produced, the program has enabled significant progress to be made, while highlighting areas where further research is still required, notably in the fields of trust in systems, learning data and human interaction.