How does AI work in Yandex Zen?

Question

Robotics

Good day!
I was very interested in the Yandex Zen project, which is controlled by AI. I am thinking of implementing something similar in my project, and therefore it is very curious how exactly the work of AI is organized in this project? I would be glad if someone can explain in an accessible way or throw links to resources on the topic of their robots and bots.

Answer 1

I was very interested in the Yandex Zen project, which is controlled by AI.

Where is infa from? At the first contact, they ask you to choose several topics of interest. Then they give a sample of them, each material falls into several topics, the more intersections with interests, the better. I don't see AI here. But to determine the subject of the article, algorithms based on neural networks can be used, which is also not AI.
For the work of AI or a neural network, you need resources, it makes no sense to fire them with a frequent event (selection for the user), it is better to fire resources when parsing an article, assigning topics, tags, etc. to it. And feed the user with a primitive selection based on standardized interests + cache, since your interests are clearly similar to another 2,000,000 people, since there are not many options for topics.
Of course, there is also a polling station, but there is also more primitive in the form of excluding tags and / or topics. The whole raisin is in the parsing of the article and its cataloging.
Although Yandex says that this is the latest achievement of human thought. They just sell marketing products based on that. That's the whole secret. Who actually uses it? I've never been given a good sample.