Welcome to LaBRI.IA, the LaBRI research Axis on AI

The LaBR.IA axis, the first to be created within the LaBRI, brings together some fifty researchers, teacher-researchers and PhD students from the laboratory’s various departments. This diversity provides a wealth of AI topics. These include application approaches such as program synthesis, HPC for AI or AI for health, as well as a range of themes that are the hallmark of the axis:

  • Processing imperfect data: The data available holds enormous potential, but it is difficult to rely on relevant and reliable information. Data comes from multiple sources, compiled differently and with varying degrees of care.

  • Aiming for trust: Despite extraordinary progress in many areas of AI, very little work imposes trust as the ultimate goal. This problem is a real brake on the deployment of AI when error matters.

  • Explaining and understanding models: Once a neural network has been trained, it may base its decisions on irrelevant elements. To improve confidence in learned models, we propose methods for visualizing the relevant elements of the image or data used to make a given decision.


Laurent Simon and Akka Zemmari