I was part of the team that built exactly this. It launched in 2010. Some Googlers of that era are probably still annoyed at all the internal advertising we did to get people to seed the data. This is one of the launch announcements:
https://maps.googleblog.com/2010/11/discover-yours-local-rec...> Google Maps shows you what the average person thinks is a good restaurant
I'm fairly sure this isn't true. At least, I still get (notably better) results searching while signed in. Couldn't tell you what the mechanism for that is these days, though. But at least back in 2010, the personalization layer was wired into ranking. You can see in the screenshots how we surfaced justifications for the rankings as well.
Pretty much immediately after launch, Google+ took over the company, the entire social network we had was made obsolete because it didn't require Real Names(tm), and a number of people who objected (including me) took down all our pseudonymous reviews. Most of the team got split off into various other projects, many in support of Google+. As best as I can tell the product was almost immediately put into maintenance mode, or at least headcount for it plummeted like 90%. Half of my local team ended up founding Niantic, later much better known for making Pokemon Go.
As for why collaborative filtering didn't take off, I can offer a few reasons. One is that honestly, the vast majority of people don't rate enough things to be able to get a lot of signal out of it. Internally we had great coverage in SF, London, New York, Tokyo, and Zurich since Geo had teams in all those places and we pushed hard to get people to rate everything, but it dropped off in a hurry elsewhere. The data eventually fills up, but it takes a while. I'm told we had 3x the volume of new reviews that Yelp had at the time, but Yelp mostly only covered the US, while Google Maps was worldwide, so density was quite low for a long time. It was probably 5-10 years before I started hearing business owners consistently talk about their Google reviews before their Yelp reviews.
Another thing is that people are really bad at using the whole rating scale. On a 1-5 scale, you'll probably find that 80% of the reviews are either 1 or 5 stars. Even more so in a real life situation where you meet the humans involved. While you can math your away around that a bit, at that point you're not getting a ton more signal than just thumbs up/down (anecdotally I've heard that's why Netflix moved away from 5 stars). And then at that point, you might be getting better signal from "were you motivated enough to rate this at all?", which is why there's the emphasis on review counts. Many people just won't review things badly unless things have gone terribly wrong. I sat in on a few UX interviews, and it was really enlightening to hear users talk about their motivations for rating things, many of which were way different than mine.