Apple’s Music Streaming Secret Weapon: Genius

I was out taking a walk this afternoon. I had my iPhone, and my headphones, and I was listening to music. When I do this, I listen to music in different ways, depending on my mood. Sometimes I pick an album and listen to it in order. Sometimes I shuffle all my songs. And sometimes, like today, I shuffle songs until I find one I like, then I create a Genius playlist from it and listen to that.

Apple’s Genius examines the music in your iTunes library and compares it with the libraries of others to create playlists of “songs that sound great together.” Essentially, Genius looks to see what you have in common with other music listeners, and crunches numbers to find the songs that will work in a playlist based on what different people have in common.

This is very different from iTunes Radio, or any algorithmically-based streaming radio, which tries to find songs that go together, sort of. It’s based on a lowest common denominator approach, trying to find bands that are similar to other bands, albums that work with a certain genre, and package them in a radio-like stream. But if you’ve listened to iTunes Radio, to stations other than those that play hits, you’ve probably been disappointed. You either hear the same songs over and over, or you hear lots of songs that you don’t like. Since you can only skip six songs per hour, per station, you’re likely to not find iTunes Radio very interesting.

If Apple makes its iTunes streaming service as I think they will, you’ll have access to every song in the iTunes Store – which is the way other streaming services work – but you’ll also have access to your own music library. The two will be commingled, and you’ll be able to create playlists using music you own together with music you stream. If so, iTunes will be able to leverage the Genius database – information about the music you already have in your library – to create much more efficient playlists and radio stations than what’s available now.

Apple’s Genius technology works quite well. I often use Genius Shuffle at home when I don’t know what to listen to, and I use Genius playlists when I’m out walking, as I described above. Naturally, I only hear music that I own, but the combinations are appropriate. If Apple can leverage this technology in it music streaming service, they may have the most personal of all streaming services, giving them a big edge in the market.

12 thoughts on “Apple’s Music Streaming Secret Weapon: Genius

  1. Comments disagreeing on the usefulness of something are annoying, but here I go anyway (sorry). This is your second post in the last little while indicating that you use Genius and I’m surprised. I have always found it to be haphazard, random, and pointless. I’d ditch it completely from iTunes if I could. I may be weird in the way I listen to my music library, though. At any given time, I’m most interested in newer things and don’t listen to tracks I’ve had for a long time very often.

    • I’ve gotten as many positive comments about Genius as negative ones, so I think it depends on the type of music in your library. For me, it works for the more mainstream music; not for the obscure stuff.

  2. I may be the only person on the planet that listens exclusively to albums, and never uses shuffle or genius-like functions. I do use playlists, but only for groups of albums.

  3. I wish Genius and predictive streaming/radio style services worked for me. I’m not the target audience for those services. I’m pretty sure those services are geared to the casual to almost moderate listener. Listeners who would believe the songs and their order served up by Genius was “genius”. The for me with Genius/Pandora etc. is all of the songs offered only differ from each other by three degrees at most. If I say I like the Beach Boys I get ‘Surfin’ USA’, Dick Dale’s ‘Misirlou’ and Jan & Dean’s ‘Surf City’ when what I wanted was ‘God Only Knows’, ‘Heroes and Villains’ and ‘Good Vibrations’. Sure, I can thumb down songs I don’t like but if I thumb down ‘Surfin’ Safari’, say don’t ever play that again, it will get swapped out for ‘Catch A Wave’. For me, Genius/Radio/Streaming style services are variety killers. Plus, with the size of the library I had it took my iPod 40 minutes to sync. Honest.

    • You’re so right — if Genius truly were genius, it would know how to separate which songs of an artist you might like and which you don’t.

      If Apple could leverage your star ratings of songs and combine that data with some intelligent parsing of a song’s audio characteristics to discover patterns in whether you tend to like songs with piano, slow ballads, songs with super fast singing, songs with tempo changes, high female voices, songs in minor keys, heavily distorted guitar, etc., etc. then they might really get somewhere.

      • I believe Genius, Pandora etc. are for people that don’t like music. They only like music in as much as it helps complete their lifestyle. Just like Starbucks caters to people who don’t really like coffee. Starbucks sells dessert in a cup, not coffee. That’s how they’re huge. iTunes is developed and improved to suit the taste of Joe (Just Above) Average and then Apple stops right there, because it doesn’t make sense financially if Joe doesn’t need it. There’s more money in phones and cameras.

        • Genius is vastly different from Pandora. It makes suggestions of music in your iTunes library, or on your iOS device. All it does is suggest music that “goes will with” other music you own, based on similarities in your music library and those of other users. Expanding it to offer suggestions that you don’t own would be powerful, since it is based on your entire music library.

  4. Emusic and LastFM, that i know of, explicitly can have users compare their own playlists or favourites. It’s still only as good as the source material, ie the playlists people have made. Saying it’s different to other algorithms doesn’t necessarily mean better, not that you, Kirk, are saying that.

    It’s a lazy way to cast a large, random net. It’s good if you have ‘common’ or ‘predicable’ music taste (i struggle with those choice of words, they’re meant in the nicest way!). If you’re looking for anything “specifically obscure” It won’t ever work.
    Add to this the APPALLING tagging that goes on, wrong artist, title, year, don’t even get me started on genre…! It’ll be a very long time before you can search for a “high energy” playlist and not get a truly mixed bunch.
    And Genius is an absurd name for what it actually manages to do…

  5. I’d like to see ‘true’ genius — one which can do a 6 degrees style matching: one which knows the lead of Dinosaur Jr. is J. Mascis, and can pick his solo material, his material for Sweet Apple and The Fog, but also the tracks on Thurston Moore’s second solo album to which he contributed guitar, or the song by The Breeders that he sang on. It could then do the same for Lou Barlow, the bassist in Dinosaur, who recorded as himself, as Sentridoh, and in Sebadoh. Then spread out further from there and make a truly related list such as a knowledgable music nerd might recommend.

    It would have to be cleverer than that for some music, especially jazz – six degrees from Miles Davis would give you hundreds of musicians, but maybe then it would filter by the areas you like. Fan of In A Silent Way? Have some Joe Zawinul, some acoustic John McLaughlin, etc. but skip Mahavishnu and Weather Report as they’re too bombastic. It’d need to know that lovers of Kind of Blue are more likely to respond to Cannonball Adderley’s Something’ Else sessions and Bill Evans at the Village Vanguard than they are Interstellar Space by John Coltrane.

    Still, I think it could be done with a clever database of artists, performers, writers and producers combined with clever tagging to describe the music.

  6. I think Apple’s purchase of Beats (with its emphasis on human-powered curation) is a sign that Apple doesn’t have great confidence in Genius’s algorithmic approach.

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.