The primary objective of WP3 is to develop an AI platform capable of providing explanations for the decisions made by AI systems. This work package focuses on understanding the various trade-offs impacting AI behaviour during the training phase, particularly the balance between data quality and privacy. Additionally, WP3 examines the wide range of configurations in which AI systems can operate during runtime. Specifically, it investigates the effects on the machine and deep learning models when trained using decentralized and distributed architectures, such as federated learning on the edge.
In our conversation with Huber Flores, WP3 leader (University of Tartu), we gained valuable insights into the main aims and objectives of WP3. We talked about the advancements made thus far, specifically addressing the tasks completed and how they fit with other activities of the project.