There are tons of "product review" videos that are little more than an automated voice reading scraped product descriptions along with a slideshow of images from a web search for the item. They have nearly no value but for products with few video reviews, view count is not a useful metric. These faux review videos get downvoted and it serves as a good signal that the video isn't worth watching.

The people who make these automated videos aren't looking for millions of hits per video, they're looking for a few hundred, which when multiplied by hundreds of thousands of this type of video, means millions of clicks for them. Since actual reviews of older products have a small number of views, the automated ones float to the top of search results. Dislikes was a way of separating the two. Few people click Like for the good videos but there were enough clicking Dislike on the shovelware videos to help filter them out. Expect to see an increase in these automated garbage videos as they're now allowed to drown out actual content.

In my experience the information getting across is just a pulse check of how others feel about the content. Tells you something about what to expect from the video.

I think recommendation centric architecture that YouTube has changes meaning of upvotes/downvotes from a global content value judgement into a clustering signal. The platform can correlate that signal to the content to cluster users, creators, content, into groups and cohorts. Those characterizations can be further used to optimize platform for ad clicks.