The use of AI tools in metrology opens new field of investigation but also brings significative challenges. AI can be used at different stage of the measurement chain, leveraging the possibility to analyse and process big data to deliver a measure, combining several input data from sensors, sensors network and sensors complex set-ups. However, AI systems often operate as "black boxes," raising critical questions about the explicability of their outputs, the risk of hallucinations, and the propagation of measurement uncertainties.
While classical measurement methods are based on physical principles and comprehensive model, and deliver quantity measured physically, AI-powered devices deliver predictions. Paradigm changes: data become the representation of the world!

Measurement data are the ground base of the AI modelling. The quality of these data, a must to deliver accurate model prediction. Beside data format, data quality from the metrology perspective consists in measurement uncertainty. Indeed, measurement uncertainty will propagate through AI model as it propagates through classical modelling, and might lead to low level of confidence in the model output. Eventually, the question rises whether the validation and certification of measurement method using AI algorithms has to be reviewed with respect to actual standards.

This round table will bring together metrology experts and practitioners to explore these challenges and opportunities through concrete industrial examples, fostering a collective reflection on how our community can shape the future of AI-driven measurement.