According to Netflix, this method is called “transfer learning”, where it takes what it learns from a source and applies it to the target to try and improve on it. In a way, it’s like cooking where you take a dish that you enjoy and think of ways to try and make it better. In Netflix’s case, they try to compare their shows with other titles that are supposed to be similar to try and determine the success and what kind of viewership they might be able to expect from it.
The AI also uses a “similarity map” to help determine what other shows in Netflix’s catalogue are similar that you might also enjoy watching. By using AI, Netflix can quickly determine important metrics, like if the AI can predict a region where a show might do particularly well in, they might start preparing things like dubs and subs for that region ahead of others and focus their marketing in that region.
It sounds very clinical, although we suppose to a certain extent it works given how many people are subscribed to the platform. That being said, Netflix does have a challenger on their hands with Disney+, who despite being late to the game, is quickly catching up in terms of subscribers.