The question is if they dont' collect feedback they don't know how many users are affected. If you break feature X and all users are affected but you send them to read some FAQ and not record the reports then the only way you get informed about this is from Hacker News or a news paper, maybe this is the Google and Apple strategy, if it causes an issue that reaches a news paper then we investigate it, otherwise we fix only the bugs that upset our developers or bosses
Most of their feedback probably comes from their metrics. Things like sudden drop in the number of purchases, an unexpected increase in refinenment of certain queries, increased number of 500s, etc. In order to respond in scale you have to try to detect unusual patterns showing up in your data. That said, they do have some avenues to contacting them. There are feedback buttons in the knowledge panel and in the translations. There's an actual support in youtube (albeit it will take a while until you get to a real person) and so on. As it is, they are likely already getting more feedback than they can process on a daily basis. Bear in mind that not all feedback is useful. A big chunk of the work in any support hotline is wading through the mountain of useless feedback. I can only imagine how that'd be like at google's scale.
Note: by useless feedback I don't mean that the user is wrong because they don't understand the interface, when it's actually the interface that's not intuitive. I'm saying that people are often calling in trying to solve something completely out of your control. Just talk to anyone in support and they will tell you the craziest things that pop up. One example I've heard was a user who could not wrap around the idea that an app on the phone needs internet access to order food delivery. Something like that can easily take 30 minutes for a supporter to handle, for instance.
