The European community has launched a call for tenders dedicated to Trustworthy Artificial Intelligence (AI). Among the handful of projects selected is the one launched and led by TRT: ULTIMATE (mUlti-level Trustworthiness to IMprove the Adoption of hybrid arTificial intelligencE).
ULTIMATE addresses the subject of Trustworthy AI trought the hybrid AI lifecycle by emphasizing the stages of building and evaluaating AI algorithms in order to deliver an operational system equipped with trustworthiness.
The principle of hybrid AI is to combine the best of both worlds covered by data-driven AI algorithms (e.g. neural networks) and model-driven AI algorithms which are based for instance on physical laws. This fusion of methods makes it possible in particular to better contextualize the data, which are it possible in particulary improves the quality of the learning process and consequently the behaviour of AI algorithms.
Even though data-driven AI and model-driven AI clearly complement each other and form a critical foundation for the adoption of AI solutions in industry, hybrid AI does not fully address the issues. Indeed, trustworthiness can be defined through the 4 axes of validity, transparency, explainability and ethics.
the objective of ULTIMATE is to pave the way for the development of trustworthy hybrid AI that meets industrial quality criteria, in tree stages :
