With the use cases listed below you learn how to design and engineer systems of Artificial Intelligence (AI) with guaranteed levels of stability, explicability, effectivity and respect for individual freedom of humans within the Artifical Intelligence Systems Engineering Laboratory (AISEL). This requires a synthesis of the best practices in modern machine learning practice with classical systems engineering methods. The design process is governed by the principles of causal inference, formulating assumptions on different viewpoints, the ability to adapt to changing requirements and validation. Graphical models with a special role of in-variances are combined with the elements sensing, acting, memory, hypothesis generation and testing. A guideline on how to reduce the space of hardware and software design configurations and are combining the best of the two worlds of deep neural networks and probabilistic modeling in hybrid approaches is given. Special emphasis is given to the interplay of causal inference and end-to-end pipelines powered by deep learning. Performance requirements and especially the time to action will be a key factor for the choice of the architecture. We demonstrate that the designed AI system can be extended in a scalable manner with the complexity of the changing requirements in context, task and performance . Simulation does play a major role in accesing design parameters and the choices for validation and modular improvements.
The use cases below contain dedicated data and requirements to allow for going through the full engineering process.
AISEL Use Cases:
