Of course, there will always be points where my idea of good and someone else's don't jive, but often there are universal quality judgements.

Ahh. Subjective analysis is a portion of our hypothetical equation. Right now it's just the AllMusicGuide rating (1-5 stars.) That might be enough of a subjective "quality" judgement to help things out a little bit. And if you had a penchant for a certain low-rated artist the idea is that your repeated high ratings of that particular artist would outweigh the one-time subjective rating which is just used to provide the initial guess as to how much you like it.

There you go. And my point, ultimately, is that some people care about quality and others seem not to, so it would be important to note whether that criterion is important to the current user or not, or, possibly, to what extent.

10 years worth of new technology, car electronics, miniaturization... what would be cool to have for ICE?

1) Each person could have a bluetooth earpiece streaming different audio rather than fight over the remote.
2) Throw video in there too... each of the passengers gets a video screen and can watch their own video stream with TiVo like controls. Wireless networking will have advanced enough that you can get your live video from home or direct from satellite.
3) Automatically switch to live radio when one of your keywords comes up or when an emergency broadcast is sent out.
4) Integrate with nav system:
a) GPS system gets real-time updates of contruction zones, lane closures, accidents and traffic jams.
b) Voice recog. "Car: I'm going to the grocery store". It may reply with "Take Duddly Rd because there's construction on your normal route".
c) When approaching an intersection, if you signal, it tells the traffic light which way you're turning. This helps optimize traffic flow in certain situations (e.g. you're turing right on red and there isn't much cross-traffic, the light knows that it doesn't need to change to give you the green).
d) As you approach your house, the garage door opens.
5) When your cell phone rings, it tells you (verbally) who's calling. You can respond verbally and tell it to answer, ignore, take a message, etc.
6) It has a combo IR/weight sensor (or something better) and learns each person's prefs. Then based on who's in the car, it picks the right Wendy Filter.
7) It can hear that you're straining to have a conversation with your passenger and automatically lowers the volume a bit.
8) The transmitter at the drive-up (bank, fast food, pharmacy, etc) tells it to lower the volume or pause (at the user's pref - or automatically quieter for music and paused for books on tape).
9) Auto-pauses when your car is out of control or in an accident.
10) OBD-IV connection detects a leaky tire and uses GPS to give you directions to nearest tire shop.

There you go. And my point, ultimately, is that some people care about quality and others seem not to, so it would be important to note whether that criterion is important to the current user or not, or, possibly, to what extent.

We've been going back and forth on that. The whole question of ease-of-use comes into play. The hope is that the weights could auto-correct themselves based on how successful they are. Like, if a particular song scores high based on quality rating, and the user gives it a 1, they probably don't agree with the quality assessment. Then on another, if their rating is more in line with what the system predicted, the weight of that factor would increase. This type of stuff makes my head spin, especially when you consider that most attributes aren't in just one dimension like the "quality rating" is. But we're doing what we can, and as long as we can B.S. our way through it, I'll have my Masters degree, and thus, have my weeknights and weekends back. Hoo hah.

Yeah these are all cool things but none would happen on an aftermarket system. These are all things that would show up in a fully-integrated OEM system.

Of course, there will always be points where my idea of good and someone else's don't jive, but often there are universal quality judgements.

Hmmm... gives me an idea. One of the things that TiVo could select on but doesn't is the "star rating" on the movies -- you know, one star, two stars, three stars, four stars. I don't know who decides what these ratings should be for a movie, but pretty consistently a three star movie is better than a one star movie, no matter what the genre.

tanstaafl.

gives me an idea. One of the things that TiVo could select on but doesn't is the "star rating" on the movies -- you know, one star, two stars, three stars, four stars. I don't know who decides what these ratings should be for a movie, but pretty consistently a three star movie is better than a one star movie, no matter what the genre.

ReplayTV will let you record by star rating, but all their 4 star movies seem to be "classics". Personally I have no interest in watching Gone With The Wind again. If you step down to 2 or 3 stars then you just get a ton of everything. The feature I would most like to see in the replay (and one that might get me to but a Tivo if they did it first) is a date range combined with the other selectors. So you could do something like record all Action-Comedies made since 1990. I can't be the only person in the world that just doesn't watch old movies and gets sick of my PVR spending half it's time recording stuff I would never watch.

Hmm, that turned into a bit of a rant

-Mike

It looks like another company is going to sell PhatBoxes. This time it's a little company called VW/Audi. From the picture it looks like it's just a different shell again. I wonder if it will have Keg in the name.

Available this spring? in VW time, that must mean winter 04. That part number has been in the computer for a while, but there hasn't been a memo to ANY department at EITHER franchise.

But one will be finding it's way into my demo as soon as they do show up.

Since you're on the subject, I'd like to point out that MusicMatch uses a very sophisticated recommendation engine to determine relationships between artists, albums and songs. We do this through anonymous play log uploading.

If a user is opted-in, we upload their MP3 playlists to the server. The server then compiles lists of similar artists based on what is in the playlists.

Basically, you can assume that the 'typical' listener likes a particular style of music and tends to only listen to that. You can also assume that they will listen to similar types of artists back-to-back. So, if a user listens to Pearl Jam and then Temple of the Dog, you can 'guess' that a relationship exists between these two artists.

Now, this process is definitely not perfect. What if you get an eclectic listener (like most of the people on this board) who listens to Pearl Jam, then Bach. Or, as you specified, everybody has different tastes and they don't always 'jive'.

Well, as we all know, in small groups, deviations will be much more apparent. But, when you take into account the sheer amount of play log data we get (millions a month), these deviations become smaller and less noticeable. This is one of the key reasons why our data is pretty good.

For example, the top recommendations for Pearl Jam are:
Soundgarden
Stone Temple Pilots
Temple of the Dog

Those are pretty dead on. Now, admittedly, recommendations for obscure artists won't be as good because we won't have as many play events for them.

But, if you're curious to check it out, here's a link to our Guide, which has a search box that you can enter an artist (and optionally, an album) and get instant recommendations:

http://www.mmguide.musicmatch.com/match/best.cgi

Actually, I kind of disagree about the connection between Pearl Jam and Soundgarden. While it may be, and probably is, that many people who listen to one listen to the other, if someone came into your record shop and asked for a band that sounded like Pearl Jam, assuming that they weren't familiar with the popular music scene of the early 90s, I don't think that Soundgarden should be the first thing out of your mouth. Pearl Jam is a big 70s-rock soundalike and Soundgarden is heavy metal, despite the fact that they both got chunked into the grunge bandwagon. I think that that's more of a statistical anomaly than an accurate musical relationship. That's kinda like saying that if I like Eminem, then I'll like Britney Spears.

IMHO.

if I like Eminem, then I'll like Britney Spears.

Although that sounds wrong, I'm sure a lot of people that like Britney Spears like Eminem (probably a majority).

Yes, just as if I like Pearl Jam, I'm likely to like Soundgarden.

But my point is that both of those judgements are based on coincidence, not musical similarity. Fifty years from now, someone might be shuffling through the oldies section of their local Ogg Vorbis store and pick up some Pearl Jam on a hunch and decide that they like the band. The chances that they would also like Soundgarden are not, I think any more significant than if they might like Iron Maiden, based solely on the fact that they like Pearl Jam.

The reason that a lot of people like Soundgarden and Pearl Jam is because they were part of the popular music culture at the time in which both bands were popular. Outside that coincidence, they bear little resemblance, and a suggestion to one based on the other is likely either to elicit either an ``I already know them'', which doesn't help, or to suggest another artist which has little similarity.

Replacing those two artists with Eminem and Britney Spears just exaggerates the issue.

Say your mom is playing on the radio and hears a Britney song and decides she likes it (for whatever reason). Is she likely to also like Eminem? I doubt it. The only people who like both like them because they're currently popular at the same time, not because they have any relevance to each other.

The point of a music suggestion algorithm is to provide suggestions to people based on what they currently like, not on what happened to have been popular at the same time. The only people to whom that would be important are the people who are already aware of both artists, and to whom, necessarily, such a suggestion would be irrelevant and pointless.

Sorry to say the same thing over and over. I can't quite figure out how to put it in words properly.

I understand your point fully right now and I agree.

In time, the musicmatch database should reflect your point (I think). But, by that time, while the differnece between Pearl Jam and Soundgarden might be reflected, the new bands will be back in the old situation of Pearl Jam and Soundgarden.

Well, the goal of these systems isn't to be perfect... You'll never match everyone's tastes. All they're trying to do is give you recommendations and suggestions that are statistically going to make more "hits" (which equals more purchases if you decide you like it) and correlating your collection with others and using their preferences to guide your recommendations is one of the better ways to do this. There are several projects I've found which attempt to match songs by "musical similarity" (including this one) but I would argue that using ONLY this kind of metric (and excluding population-based data) would tend to limit one's exposure to new music which may be somehow relevant to songs they already own, but might not necessarily "sound" like it. While Soundgarden and Pearl Jam might not correlate highly via a musical similarity algorithm, the fact that you see them both in a lot of peoples' collections *should* be a factor in any recommendation engine... Just not the only factor.

Right now my project team and I have four basic factors that are going to go into the equation which will rank songs for new music recommendation. Population-based data like MusicMatch is using is one of those factors, and right now I think we're leaning towards it being 30% of the equation by default. But we've got seven weeks to figure out if that's good or not, and really, since we won't end up with a legitimate working system, it's all a bit hypothetical. But if I were developing a real recommendation engine, that's about as high as I'd go with the use of population-based data.

I own a riocar and a PhatBox. The riocar wins hands down, however there are a couple of advantages to the PhatBox. It will work with a few factory head units, and it installs in about a half hour. If you do not have a factory head unit that is compatible, then it has NO advantages.