We think that intelligence augmentation systems should be based on a type of artificial intelligence that goes beyond pattern recognition (i.e., deep learning, …)
Instead, we propose to develop systems that build causal models of the engineering systems of interest, able to provide explanation and understanding to the physical phenomena taking place.
To achieve this ambitious goal, we propose to employ different ingredients:
- Hybrid Twins. Since the developed system will obtain information from the surrounding environment through computer vision, it will posses valuable information (data) to assimilate, but also possibly to correct those models that may be incorrect or incomplete. This fits into ESI’s vision of what a Hybrid Twin is: a dynamic, data-driven twin that is both physical and digital.
- Real-time simulation. For this to be possible, one bottleneck is our ability to perform high-fidelity simulation in (or faster than) real time. This will be possible by employing model order reduction techniques and, particularly, by exploiting our vast knowledge on Proper Generalized Decompositions, acquired through our longstanding collaboration with prof. Chinesta and his group.
- Augmented reality. Finally, information arising from the embedded models and the acquired data will be displayed by means of augmented reality devices (tablet, hololens glasses, …)