UNIT.City — місце, де люди працюють... КРАЩЕ! Обирай свій простір просто зараз 👉
Олег ОнопрієнкоAI Eng
8 June 2026, 16:33
2026-06-08
Feature engineering, auditability and ensembles: The Ministry of Digital Affairs has updated the official dictionary of AI terms
The Ministry of Digital Transformation of Ukraine has presented an updated edition of the dictionary of terms in the field of artificial intelligence. The document has been supplemented with 40 new concepts that reflect the latest technological trends and international regulatory practices.
The Ministry of Digital Transformation of Ukraine has presented an updated edition of the dictionary of terms in the field of artificial intelligence. The document has been supplemented with 40 new concepts that reflect the latest technological trends and international regulatory practices.
As explained in the department, the goal of the initiative is to help the state, business, media, and researchers speak the same language. The presence of clear definitions, each of which has an English equivalent, will allow specialists to interpret processes in the same way during the development of AI products and the preparation of the regulatory framework.
In addition to highly specialized technical terms, the updated list explains a number of phenomena and mechanisms relevant to the industry. Among the new concepts:
AI-slop is low-quality mass content from generative AI, where speed of replication prevails over accuracy and content.
Query injection is the manipulation of a language model using specifically formulated texts.
Model drift is the loss of accuracy of a model due to changes in real data over time.
Superintelligence is AI systems capable of self-improvement that significantly surpass human cognitive capabilities in all areas.
System request (initial context) — a starting instruction for AI before interacting with the user, which sets the rules and style of work for the entire session.
For ease of integration into workflows, the dictionary is available not only as a text document, but also in a machine-readable format. This solution is aimed at developers, researchers, and organizations working with large data sets and neural network training.