Collecting IT specifications of good practices in AI

The work presents the study of specifications of good practices in applied artificial intelligence (AAI). The analysis of 25 questionnaires from five partner institutions revealed key insights into the current state of artificial intelligence (AI) and machine learning (ML) projects. Training conducted in Serbia and Bulgaria, was signaling a need for expanded opportunities in EU countries. As a result of the study, we obtained that Deep ML prevails, particularly in Convolutional Neural Networks, while Gated Recurrent Unit is less common. Data volumes between 1 GB and 1 TB are typical, reflecting practical constraints. AI applications span diverse fields, with TensorFlow leading in libraries. Permissive licenses are most prevalent, databases are primary data sources, and texts/pictures dominate data characteristics. NoSQL databases are favored for storage. Security features and data processing tools vary. Dedicated servers and clusters are widely used, recommender systems are prominent, Python is the preferred language, and Apache Hadoop dominates ecosystems. Free datasets foster accessibility. Overall, the findings emphasize the dynamic nature of AI/ML projects, providing a foundation for future research in the rapidly advancing field.

Full paper can be found here: Research 7 StateOfTheArt_V3