Hacker News Reader

Comment by Pooge | Hacker News Reader

I know this is an anecdote and very subjective but I've really never discovered something that I loved with a recommendation algorithm.

Whether it be videos, music, livestreams, books... Everything that I've considered a "10/10" has been recommended by a human or a non-personalized algorithm—such as "Most popular". Whether that's a direct recommendation by a friend, a comment on HackerNews, someone that I already follow that mentioned the thing in question.

My RSS client fetches my YouTube "subscriptions" and it's been years since I've been on the homepage.

Surely I'm not alone but it surely feels like it.

View on HN

I’ve had some great recommendations from YouTube. For example the video below, called “The Squirrel and the Peanut”, which by just adding some well chosen classical music to some found footage of a squirrel getting a peanut turns it into a moving story of overcoming your fears told in a mere 1:32.

Watch with the sound turned up.

https://www.youtube.com/watch?v=LQ1ZYGHmtN8

I agree that the algorithm isn't always the best. Interestingly however I occasionally get "legendary pulls" (I think that's a gatcha term?) on youtube - a 1990s stop motion video of some legendary niche Japanese animator, or some other 8+ years old video in 240p that's absolutely amazing.

But "fortunately" for me, on my main account, my feed is 80%+ decent stuff - science channels, AOPA safety videos - stuff like that. But my other account has absolutely terrible suggestions on the algorithm, so it probably varies a ton on how much and what you watch.

> Everything that I've considered a "10/10" has been recommended by a human

Have you ever found a 10/10 on your own?

If so it's possible you were recommended that by an algorithm but you just didn't register it, because a human recommendation is more of a memorable event.

> Have you ever found a 10/10 on your own?

If I've ever did, it's after having been recommended something. For example if someone sends me a song and then I go to the "most popular" songs from that same artist. Sometimes I would stumble upon something I like even more than the recommendation.

For example if it's a genre that interests me I would search "[genre] mix" on YouTube. In which case it's usually mixed by a human.

Akshually I disagree, I discovered the song "Plastic Love" from a YouTube recommendation. Then I dived into city pop as a result.

But taking into account all the thousands of not-so-good recommendations I guess you are right

Most of the great music I've discovered is thanks to Spotify's "Play song radio" feature.

Granted I have very specific tastes in music and 98% of the time I listen to things I've already listened to hundreds of times... I sometimes try the song radio but it's not to discover a "banger".

As much as I dislike Spotify, I've definitely discovered some 10/10 tracks or artists that I would've been unlikely to have found out about otherwise. I don't know how well it still works these days since I switched to Qobuz years ago (which is great in everything except personal recommendations), but Spotify's algorithms, especially for Discover Weekly, used to be amazing. There was an article about all the kinds of stuff they do, and it included things like recommending less frequently played tracks with acoustic similarities to your favourites, and tracks presented on music blogs that also featured music you like. The stuff the engineers got to play around with there before the service started getting enshittified, basically.

Reply on news.ycombinator.com