Developers of a domestic voice robot for management companies in the housing and communal services sector faced an unexpected problem — their neural network learned Russian profanity in just one month of talking with clients. This was announced by Mikhail Viktorov, President of the National Association of Organizations in the field of Information Modeling Technologies (NOTIM).
Literally in the first month, the developers noted such a collision that the neural network learned to swear. Therefore, this collision, of course, had to be eliminated. But nevertheless, these are indicators of active work with our citizens.
The use of voice robots allows reducing the staff of call centers of management companies by 5–6 times. Instead of 20 employees, only 2–3 people can work, since the neural network processes 80–90% of all requests.
Mikhail Viktorov emphasized that, according to statistics, 80% of clients receive all the necessary information from the voice assistant. Dissatisfaction of citizens is noted only in emergency situations. In such cases, it is necessary to switch to a live operator who can make emergency decisions.
Read also materials on the topic:
- AI was taught to automatically filter unwanted content in video or audio materials
- Recognizes jokes, swearing and even sarcasm: Yandex Cloud introduced a new Empath neural network that distinguishes emotions in speech
- Digital housing and communal services with computer vision and speaking AI appeared in the Nizhny Novgorod region
