Slashdot Mirror
Is the iPod Shuffle Playing Favorites?
marksilverman writes "Steven Levy at Newsweek is reporting that his iPod Shuffle seems to favor certain songs. Is Apple receiving kickbacks to promote certain artists? Apple denies it, of course, and Levy had the good sense to ask a mathmatician and a cryptographer who explained that it's probably just humans finding patterns where there are none." Less neurotically, both CNet and PCWorld have discussions of the Shuffle's interior spaces.
Computers can't generate true random numbers (ok, at least I don't know of any current methods) but only pseudo random numbers. There's a precise mathematical description that gets you from one number to the next.
Who knows, maybe Apple uses the meta data for a song to determine the random order (anyone hack it yet and finding the algo?) and some people just get "lucky" like prof John Allen Paulos explained in the article. You might happen to flip 6 heads in a row (despite being a 1 in 64 chance of it happening) and you might get 6 songs from an artist in a row.
Sounds to me that it's a conspiracy theory at best.
:wq
If you have any audiobook or spoken word that have proper metadata, they will never be selected in shuffle mode. While this can be handy for not falling onto a 20 minute chapter of a book randomly, it also makes it a bit more difficult to create cut-ups, or experiment with random spoken words when you want to: you must retag the tracks.
This is also useful to take long tracks out of random selection. A friend retagged Pink Floyd's 23 minute long 'Echos' as a book after getting pissed off that his Shuffle always seemed to select it.
When looking at the candies through the side of the glass jar, the first thing you notice is that the distribution of red and green candies doesn't look evenly distributed at all. Instead, there are lots of areas where many red candies are adjacent, and lots of areas where many green candies are adjacent.
For a long time, many people thought there must be some kind of static electrical effect present that was causing candies of the same color to tend to stick together. Eventually, however, some statisticians did the math and found that there was no such effect at play -- in a completely random system, such "blobs" of like colors are inevitable. Indeed, a jar of candy with no such blobs would be a bit suspect -- what are the chances of the red and green candies always pairing up so that no groupings occur?
To put it another way: it's all in your heads, guys.
I don't care if it's 90,000 hectares. That lake was not my doing.
The odds of the iPod playing a completely even distribution of every song, with no detectable 'pattern', is MUCH smaller than it playing a list that has a detectable 'pattern'. Think about it this way.
Songs 1, 2, 3, 4, 5.
And you want to play 5 songs. To have it play each song once and only once would be 5!. However, this is out of the number of choices total, which is 5^5.
In other words, it's 120/3125 that the distribution would be even. As opposed to 3005/3125 that it wouldn't.
There are some quality random number generators on the Internet like Random.org, HotBits, and Lavarnd. But to be technical, their numbers come from background radio noise, radioactive decay, and lava lamps (er, "Lava Lite lamps"), meaning they don't truly generate it on the chip.
This sig is certified free of self-referential humour!
I noticed that screens towards the top end of the scale seemed to show up more than screens towards the bottom (ie. if you had twenty images, images one to ten would should up more than eleven to twenty). I did some sort of multipass algorithm to stop this happening - I forget exactly what I did now and the source is lost, but I definitely had to change things to make it acceptable.
Cheers,
Ian
Yeah, the iPods are like that.
I have this playlist with 5 albums on it. Every time I mix it up again, I can predict a result. Whatever the first song is, I will, EVERY TIME, hear 5 songs from that album before I hear 3 from any other album in the playlist.
iTunes does it too. Whatever randomizer they use, it's a little funny.
On a similar vein, all 'shuffles' in my music players are abysmal. The best I found was an application called SAM2, which is designed for webcasting, and was absolutely brilliant. No same artist within 30 minutes, no more than 3 tracks from an album in 2 hours, and it worked with ratings as well!
Perhaps Apple would care to realise that their 'star' ratings have more use than just excluding songs from playlists?
Hell, WMP recognises which songs you prefer at what times of the week. For a company so focused on perfection as Apple, it surprises me that this isn't standard in iTunes.
How many people can read hex if only you and dead people can read hex?
Actually I am guessing you hit the nail on the head.
.01, .02, .03, .04, ... .98, .99 and there are only a hundred of those, so there were only 100 possible choices spread across the 300 people on the list, and those 100 choices were coming up a LOT (no surprise there.)
I was once brought in to fix an application that randomly picked people to pee in a bottle, and a few dozen people complained that they got picked a lot more than other people (out of about 300.) Well it seems that the 'programmer' that coded the app used two digit precision on all math calls (the coder was generally assigned to fiscal apps, two decimal places in money = two decimal places in his head.)
For those not reading between the lines, a random number returns a real number somewhere between 0 and 1, this real number is then multiplied by the maximum number in your expected range to come up with a random number between 0 and your top number (300 in this instance) but since he was using 2 digits of precision (ie, two digits after the decimal place) all the random function returned was
I figure it has something to do with the random number generator in the baby-iPod.
Glonoinha the MebiByte Slayer
Worse than that. Many implementations of rand() have extremely un-random low-order bits, so the above code will produce both predictable and unevenly weighted selection. It's much better to do
const int song_id = (int)((random()*(long long)num_songs)/RAND_MAX);
(long long cast required because max songs * RAND_MAX might overflow an int).
One way that I deal with this in iTunes is to make a smart playlist that only includes songs I haven't listened to in the last X days.
:)
Unfortunately, if you fast forward through a song, iTunes doesn't count it as played. This means songs you dislike will pop up even if they've popped up in the last X days.
So I make the smart playlist include only songs rated 3 stars or more; when a song I don't like pops up, I rate it as 2 or 1 stars and it stops showing up again.
Works as well as anything that doesn't include telepathy and AI to predict exactly what I want to hear at a given moment
On macslash.org, there was an Ask MacSlash about iTunes somehow figuring out what songs sounded good together using some crazy sonic algorithm. The guy was asking if it was plausible that iTunes analyzed the songs in the playlist to spit out the best mix possible.
Well, there is ways of analysing the song to figure out what type of song it is -- There is actually a very intelligent professor at my local University that has developed an algorithm/program that can tell with a pretty good idea what type of song it is. (that might not be a correct link, I was poring over their website to try and find any information about the software/visiting professor and I can't say that I recall all the details exactly.. =\ ) I saw a demo of it in a class one time when he was a guest lecturer and it was pretty impressive.
Not to mention that DJ thing you can do with a phone.
in girum imus nocte et consumimur igni