There are two different subjects that I want to touch on re:hippies. I just had to get that first post out to lay the groundwork for a larger discussion on (1) social boundaries/PCorrectness and (2) whether patterns or metadata could be used to reconstruct personality. I think I'll start with the latter.
I used my love/hate affair with hippies to highlight the peculiar human dilemma (as far as we know it is only our burden) of having multiple emotional responses to a macro thing without always being able to understand (on the micro level) what discrete pieces we are responding to. In other words, if I were able to truly break down my personal understanding of "hippie", could I ever really separate out what I hate from what I love? Does my mind have the capacity to see and understand trends and patterns, or would a computer be much quicker, potentially smarter, than I at this task?
I have been reading a lot lately about Web 3.0 and the Semantic web. One of the most intriguing aspects of the semantic web is the effort to use metadata to connect and teach computers (or a larger AI) to recognize what is being represented and to not just simply represent it upon command. In other words, there is a large-scale push to make it possible for computers to not just display a graphical map but to understand what is being displayed and relative connections that might exist within it.
umm...at least that is my personal understanding of the terms today. I'm not quite sure if I "get it" yet...
So...I wonder if my personal metadata could be understood by a force more powerful than myself (AI), and, if so, could algorithms and connections be used to (1) understand me and (2) reconstruct me (not physically, of course, but virtually).
I have to begin with music and my personal musical tastes since music is prepacked with metadata, and it is easy for us to understand what exists in that space. So, given enough data about my musical likes, dislikes, and indifferences, could the metadata of said music lead to an actual understanding (and accurate prediction) of what my musical-tastes profile is?
So, here is a random list of some of the songs on my current playlist that I love (truly random; just hitting shuffle and listing the first five songs):
1) Brokedown Palace by the Grateful Dead
2) Shooting Shark by the Blue Oyster Cult
3) Homesick by Kings of Convenience
4) A Good Country Mile by Kevn Kinney
5) Said I Loved You...by Michael Bolton (oooohh...this one hurts to admit. bad shuffle luck).
Granted, way too small of a dataset, but you really don't want me to list all 64 songs (and I don't want to admit to some of them).
Let's look at the metadata and patterns that could emerge:
Metadata
1) Song title
2) Artist
3) Producers (and the like)
4) Date of publication
5) etc.
musical patterns
1) rhythms
2) chord progressions
3) harmonies
4) instances of diverting from internally-established patterns
5) etc.
"non-musical" patterns (admittedly a misnomer)
1) lyrical (though it will be quite some time before a computer will understand "you drove me to the wall")
2) linguistic (frequency of interdental sounds, fricative, etc.)
3) voice tonality
4) rhetorical patterns
5) etc.
And what about other characteristics of every particular song: levels (how loud/soft is voice compared to instruments), blending, background noise, etc.
I am starting to think that music might have not been the simplest place to start this discussion... :o) At this rate, you would think it matters whether the singer was facing east or west (and perhaps it does).
Back to the larger question. Could a program analyze and gather data from a representative set of my musical likes and dislikes and understand my taste? I certainly don't understand the connection between loving Billy Joel's "The Downeaster "Alexa'" , Vengeance Rising's "Fill this place with Blood" and Patty Griffin's "Rain." And I think I would be quite miffed at a report that explained what the connections are and why I love them, essentially reducing my musical taste to a number.
However, the idea intrigues me. Is there a connection? There must be since some music resonates with me and some does not. So, what is that connection?
The problem for the AI would be having to account for non-metadata influences, like moods and memories. Of course I have emotional ties to many of my favorite songs/musicians, and it is very hard to fathom a virtual entity being able to quantify such things.
Color me interested, however, in how the semantic web might evolve beyond 3.0 and not only understand meanings and connections but perhaps even digest, dissect, report, and recreate them.
shanti,
mjh
p.s.--if some of these posts seem random, incomplete, suddenly-ended, scatterbrained, jumpy, etc., it is because they are. I am still trying to get my writing/thinking/time-management legs back. And I have a 2-year old.
6 comments:
i have to re-read this, i'm forming a real comment.
and about your p.s. - thats what a blog is about, silly. don't apologize.
Crowd-sourcing is a better way to determine musical tastes. That's what last.fm is for
I'm looking forward to re:hippies, part 3. In the meantime, rock on for admitting to the Michael Bolton. I always hate it when people look at my iPod and see my music list. Sure there's the cool Raconteurs and James and Badly Drawn Boy--but there's also some really embarrasing things in there too.
Have you tried with the genius playlist on iTunes yet? It can't predict what you'll like based on what you listen to most--but you can pick a song and (like Pandora)it picks out a bunch of suggested songs from your library. What's really cool is that get to rediscover songs that you haven't heard in a while or that's buried in an album that you usually only listen to for the first three tracks and then change it...
As for hippies, I feel your conflict.
I would love to discover how sites like last.fm, launch, imeem, etc. predict and define musical tastes. launch seemed to do a pretty good job (at least for me) back when I used it, but that has been some time ago. I wonder if last.fm would stay in-genre (i.e., if you like Patty Griffin, then you might like Dar Williams) or whether a pretty specific profile could be built based on hundreds (or thousands) of personal and crowd ranks/links. I used last.fm for a little while and probably still have an account over there. perhaps I will go check it out again. thanks, Jose. I need your smart, computer-savvy comments and direction b/c, as you can tell, i am trying to work through all of this tech-stuff with an English-major brain. Not always a good mix.
Rob, I haven't tried iTunes genius, but I could use some reintroduction to my own(ed) music library. I listen online so much now that my virtual albums are getting dusty.
---turning on Genius now
---hitting Party Shuffle
Apparently "Pets" (Porno for Pyros) goes genius-ly well with "Patience" by Guns N' Roses. Ahh....GNR.
Looks like this uses crowd-sourcing, similar to what Jose was talking about. Gotta love the masses. Wait, why am I blogging? GNR is on.
my comment was going to be about food. for example, the coupons that come out with your receipt at the register. when you buy certain food - it suggests other things that they think you'd eat.
also, rfid is a big thing. if they can track what you're buying and where you're going with it, they can find out what whole neighborhoods and towns eat.
and then i though 'who are "they"?' and i realized that "they" is me and the big G thats on my shirt right now! yikes.
Post a Comment