Towards Continuous, Autonomous, and Sustainable Artificial Intelligence

The BEREZ-IA project aims to advance knowledge and develop new methods based on bias-free Artificial Intelligence (AI), capable of autonomous learning and operating sustainably in its training and inference processes. The motivation for this initiative arises from the convergence of interests and practical limitations observed by the participating institutions and centers: current AI developments operate in controlled application environments under highly restrictive practical assumptions (no variability in data/problems, model performance being sufficient for the problem at hand, always available labeling, and virtually unlimited resources — both hardware and energy). Reality has revealed that such assumptions are not only difficult to guarantee in many application scenarios but also limit the potential of the latest AI models in terms of adaptability, autonomy, and sustainability under changing conditions.