Live data from Hacker News

Show HN: I trained an AI model on 120M+ songs from iTunes

maroofy.com

81–90 of 444 posts

Re: Show HN: I trained an AI model on 120M+ songs from iTunes

#81
post #51

Exciting to see AI used this way. My main feedback is I'd look at incorporating other factors to rank results, not purely how similar it sounds. Audiophiles might prefer a pure similarity ranking, but that could be offered as a non-default setting if anything. e.g. I'm sometimes seeing several essentially identical tracks at the top of recommendations (also mentioned in a comment by rayshan). You probably want to pen…

I think popularity based ranking should absolutely be optional and a toggle. I think it's reasonable to have default rank be popularity, but in my opinion, the value of a model like this is finding obscure tracks. I also would definitely like seeing an exposed toggle, and maybe automatically toggle it off when someone presses refresh?

Sure, a toggle is one option. The main point is to introduce popularity as a factor and ultimately the best UI and defaults are best decided through A-B testing.

I'd also add that it's usually a good idea to incorporate this setting into the URL, regardless of the UI, so that a specific order can be bookmarked and shared.

Re: Show HN: I trained an AI model on 120M+ songs from iTunes

#84
genuinely stoked it found a track that sounds like Cicatraz ESP by Mars Volta in Addiction by Tripp Berlin (which is new to me, so thanks for that!) but as others have said in other cases it fails to understand the pull of a given song but you can always understand why it recommended a given track.

One outcome I found interesting is that I gave it Ante Bijou Up but it didn't find any of the original versions of Ante Up based off of it and its first recommendation was some hideous pop song.

However, based off the success of the Cicatraz ESP recommendation I am a massive fan of your work! <3

Re: Show HN: I trained an AI model on 120M+ songs from iTunes

#85
Something I’ve dreamed about but haven’t found: a tool/service/etc. that can take my tastes from an era, say my eclectic early 2000s mix of jazz, electronic, and tango, and find me a similar set of music I might be interested in from the 2020s or the 1990s. I would love to explore my own taste in music in different eras. Interesting work. Reminded me of my little dream.

Re: Show HN: I trained an AI model on 120M+ songs from iTunes

#87
https://maroofy.com/songs/1651341589 is one of my favorite songs, but it says, "Hmm, something went wrong," and shows no similar songs.

When I was a teenager, I thought no band was as cool as Rainer Maria.

Does this mean I was right?

:D

Nice app! I tried it for another fave https://maroofy.com/songs/1578598760 and will give those a listen.

Re: Show HN: I trained an AI model on 120M+ songs from iTunes

#88
Really cool! From my understanding, it looks like this is doing something like word embeddings and searching for nearby points in the embedding space.

Crazy idea: what if you used a dimensionality reduction like t-SNE instead of learning a vector representation? Would you expect similar results?

Post reply on HN