Widespread availability of ChatGPT has posed new challenges for teaching and learning. The paper examines whether and how the chat can be used to support teaching and self-learning in engineering  and technical sciences domains. ChatGPT (version 3.5) was tested and assessed in a similar way as our students were. Different types of exams (exclusion tests, multiple choice tests, open questions) were carried out in two areas: control theory and metaheuristics. On average, the chat received quite good marks. However, it had a problem with questions that required detailed information and understanding of the subject. A number of conclusions have been drawn from the research. In general, AI can be helpful to teachers and students if used wisely. However, it must be emphasized that the use of the chat can not be uncritical, and almost all information should be carefully checked and validated, especially in engineering and technical sciences where precision and unambiguity are important.