Usage of data for the solution of direct and inverse problems in mechanics and structural analysis has been the topic of various investigations in the last decades. Complexity of neural networks has drastically increased, leading to deep learning tools, while additional options like differentiation of the neural network metamodel has facilitated the development of physics-informed, self-learning versions of them. The previously outlined set of artificial intelligence tools are currently being integrated into finite element modelling codes and are used for the creation of neural network assisted reduced order models in order to support the creation of structural digital twins. Review of current activity in the field, an academic example of direct and inverse multi-dof oscillator solved by PINNs as well challenges and difficulties for broader usage of this technology in the field of structures and civil engineering are discussed here. The availability of sensors and Internet of Things promise a digital upgrading of existing structures for efficient structural health monitoring, energy management and other tasks.