Word in technogenic multidimensional space

Today, artificial intelligence is actively mastering natural languages, becoming an interlocutor and partner of human in various aspects activity. However, the symbolic approach, which implies transfer rules logic, has failed, number exceptions language does not allow its formalization, so modern «deep learning» neural networks involves independent search for patterns extensive databases. During training, puts a word into sentence that syntagmatic relationships are as close possible to those target base, taking account both semantic words between sequence presentation. The «language» information technologies digital. processing, represented vector form numbers. idea representing mathematically familiar people usually associated with logical consistency. Visualization position multidimensional space created by demonstrates patterns, obvious syntactic relationships, but essence other obvious. mathematical representation words, intelligence, can you look at from new non-human point view.