And it is precisely on this point that Orobix, as a partner of the AI Observatory, has given its contribution during this year of activity. Orobix suggested to include a specific question in the interviews with the Companies, regarding the governance and monitoring procedures put in place to manage AI after it has been put into production, in order to understand the level of sensitivity on the issue and get a picture of the ongoing initiatives.
RESULTS OF ARTIFICIAL INTELLIGENCE RESEARCH:
It emerged that one Company out of three has not yet implemented any procedure for the management of automated systems, to ensure their safe, reliable and continuously improving use. Few companies have adopted versioning systems for models and data and procedures for their management (A/B testing, continuous deployment, tests on validation datasets). Even fewer companies have fully understood the importance of data input to the model within an artificial intelligence project and how fundamental it is to keep them under control in real time and correlate them to model performance.
The companies that have taken some actions have mostly relied on the preparation of periodic reports on model performance, while few have chosen to implement a real-time monitoring system. Finally, companies are still far from implementing active learning systems through the automatic construction of datasets for the retraining of models, aimed at the continuous improvement of models.
CONCLUSIONS: WHAT ARE WE DOING FOR ARTIFICIAL INTELLIGENCE DEPLOYMENT?
We read these results as a starting point and a confirmation of Antares Vision Group's vision on AI deployment, monitoring and governance, with the goal of ensuring compliance and reliability in all business contexts involving artificial intelligence.
In concrete terms, we are working to offer our clients the possibility of fully exploiting the potential of artificial intelligence through invariant.ai®, the Orobix platform that guarantees full observability of the AI systems put into production and of the connected processes. From model serving (even on devices with reduced computational resources), to the control of the functioning of the machines involved, to the correct execution of services and models and the improvement of performance over time. Up to the complete governance of processes, ensuring compliance with internal procedures and international industry standards.