i love “personalized radio” but i don’t listen to pandora. i like last.fm because it knows what i listen to and have listened to and programs accordingly

but i only listen to last.fm about 20% of my listening time.

i love mp3 blogs and listen to them via the hype machine all the time

i also love internet radio and listen to my favorite shows and DJs all over the country.

and i still listen to a lot of my library via playlists and whole albums

and of course, i blog a song of the day everyday at fredwilson.vc and turn it into a radio show at fredwilson.fm

i realize not everyone is as nuts about music as i am or you are, but if you love music and the web, you have to mix it up, share it, and engage with it.

And while I’m thinking about it, I trust music fanatics second. I trust user-based similar algorithms third (e.g. emusic’s). I trust friends fourth. I trust the radio last.

I like this.

Keep in mind, BTW, that I’m not talking about commercial radio — but rather, about college radio shows streamed across the web run by music fanatics. So if we can agree on that distinction, I think we’re in the same boat.

I don’t use Pandora. I casually listen to the radio, but never with the expectation that I’ll find something I love anymore (though I did do that at one time, on some stations, for some programs–but whether it’s me that’s changed or the radio, that doesn’t seem to work for me anymore).

Like Chris said I’m suspicious of algorithms that determine similarity except, like emusic’s, when they are based on some commonality in user selection. What I have learned to trust with almost full reliability is artistic connections. For example, my fanaticism for Nick Cave led me in many musical directions: Mick Harvey was in Crime and the City Solution while he was in The Birthday Party, and I love the BdayP and recognize Mick’s style, therefore I should check out Crime. Crime featured Simon Bonney, who’s voice was always spot on even when Crime’s songs were not, therefore I might like his solo work). Blixa Bargeld played on Mutiny/The Bad Seed, and he has a German band who is released on the same label (and advertised on the same page) as New Fast Automatic Daffodils, who are weird but I “get”, therefore I should check out Einsturzende Neubauten. Etc.

On the flip side, if as a college Freshman I was wild about Possum Dixon in a quirky, Possum Dixon-esque way but found out that Rick Ocasek was producing there next album I would know that I respect but am not a big fan or Rick Ocasek or The Cars and therefore might not dig his influence that album (I didn’t).

So for me capturing artistic connections is important–if we could do that better with technology, great. I think we could apply this to OER, so long as we track authorship in some meaningful way.

And while I’m thinking about it, I trust music fanatics second. I trust user-based similar algorithms third (e.g. emusic’s). I trust friends fourth. I trust the radio last.

I suspect that, for the most part, none of us could since the album no longer (generally) represents a thoughtful selection but is a collection of songs meant to be singles– which is even *worse* than Pandora!

Yes, this is a problem — and not one that was brought on us by technology, but by business consolidation. Still, I think there’s good albums out there for those into the indie scene. The Blizen Trapper Furr album is beautiful front to back, the King of Prussia EP is a genius 7-track odyssey. Vampire Weekend’s album in January was one of the best debut dics I’ve heard in ages — I’d put it next to any debut album in the last 20 years on execution of concept.

I think part of my slant is that Web 2.0 is supposed to help fix that culture of the album as collection of singles. That by subverting the A&R radio scam, we get rid of this payola induced nightmare back to stuff with some real weight. Pandora seemed to do the opposite for me, although maybe others use it in a more responsible, less brain numbing way than I.

I do find it odd that you would prefer 80% stuff you like chosen by a person than 100% chosen by algorithm. If a DJ regularly picks out music I don’t like, then I’d stop listening. At what point does that cross over for you– would 90% music you don’t like be better because it was a human selection? 80%?

I may have not been clear here: I actually *like* being presented with songs that don’t move me — as long as it’s by a DJ I trust. It forces me to open up to it, to give it a chance. I’m not sure why that is a more meaningful process with a “real” DJ, but it is. May be a mental block on my part. Maybe I hit the thumbs down on Pandora too fast. On WFMU there’s a DJ that often squeezes in a The Four Tops song into the set. And Four Tops is something I generally will flip past. But in that context, he’s built up a trust with me, and I’m ready to be open to it — and after a bit of not really getting it — I suddenly get it… does that make sense?

Once again, whether one could achieve the same results with Pandora, I don’t know. I just know that for me, that sort of experience disappeared.

On the other hand, do you really think, with most albums of the last couple of decades, that you could tell the difference between an album composed and selected by a group or musician from one that was selected from a pool of songs by a computer (or a soulless label rep)? I suspect that, for the most part, none of us could since the album no longer (generally) represents a thoughtful selection but is a collection of songs meant to be singles– which is even *worse* than Pandora!

I’m not sure I’d put your comments here in opposition to intelligent agents– there’s plenty of role for them in many areas even if this one were set aside. And in most cases they don’t need to be as smart as us or human, they just have to be smart and human-like enough. they already are smart enough to replace most of our DJs up here in the far north… but that isn’t saying much!

I do find it odd that you would prefer 80% stuff you like chosen by a person than 100% chosen by algorithm. If a DJ regularly picks out music I don’t like, then I’d stop listening. At what point does that cross over for you– would 90% music you don’t like be better because it was a human selection? 80%? Maybe, again, it’s lack of good human DJs, but I prefer to listen to what I like regardless of provenance. I’m just not that interested in most DJ’s opinions… not as much as I am interested in my friends, who I vest with much more of that kind of power…

a) there is not a large push yet for intelligent agent/OER models, and

b) there are other ways besides courses to author larger experiences out of OER

On point a, I’ll just say this is coming — at least the push for it. Intelligent Agents are a product, blog-based networks are not — that means intelligent agents will be pushed. And the message will be to cut that messy humanness out of the equation. They will ultimately lose, but the argument is on it’s way.

On point b, I’ll concede. Still, if I can enjoy an album in an age of digital download, if it can still be a meaningful unit despite it initially being a product of encoding limitations, I have no doubt that courses, whether they be two sessions or twenty, will still continue to be a useful mode of delivery.