In my opinion, the researcher, when sitting down to report on his research, should always produce new text, and avoid reusing text already written on other occasions. This should be more present in grad school.
However, what characterizes plagiarism is the literal copying of excerpts from other authors' work. So-called self-plagiarism is not characterized as plagiarism per se, because it is not possible to "copy oneself".
In times of high availability of texts, automated translations, I believe that this scourge will accompany us for a long time, and its solution depends on training, not detection tools.