Beyoncé's and Kanye West's listener numbers manipulated in Tidal
translate.google.com
translate.google.com
https://www.dn.no/staticprojects/special/2018/05/09/0600/dok...
DN is a financial newspaper in Norway, pretty important and trusted, if slanted towards "business" points of view.
NTNU is the second biggest university in Norway, focuses on science, and is well-regarded.
DN somehow aquired some files that purports to be logs of songs played in Tidal. The numbers in those logs seems to match up with other reported numbers, but Tidal denies the logs are genuine.
Researchers at NTNU analyzed the logs, and found that if they are genuine, then data was manipulated to boost play numbers for Kanye West's and and Beyonce's albums.
https://www.musicbusinessworldwide.com/did-tidal-falsify-str...
I'm surprised this isn't bigger news if it is true.
>“It’s physically impossible,” he says.
>The list goes on.
The sql query that generated all those fake listens shouldn't have passed QA. /s
What if he simply put the album on repeat and left it running somewhere by mistake? There are 20 tracks in the album for a total length of 66:39, on repeat you'll rich 96 tracks played in a little over 5 hours. It doesn't sound implausible if it plays in the background while doing chores or other things. That would explain the plays during the night as well if he left it running on a muted computer.
Not that it invalidates the rest of the study or that it would surprise me if somebody had tinkered with the numbers, I just thought that it was a bit sloppy to use "physically impossible" when it's just "unlikely".
I once turned down the volume for a streaming service one evening probably to talk to a friend. The playlist played through the night and well into the next evening. Depending on the playlist, some tracks could have gotten 15+ plays due to absent-mindedness.
I can definitely see a music critic putting together short playlists (e.g. 3 songs) for work on a specific album, article, or project. 3 songs at 3.5 minutes each in a playlist would take 10.5 minutes to play and loop 5.714 per hour. You could get 96 plays out of all 3 tracks by continuously looping the playlist 16.8 times.
Again, I'm not saying the numbers aren't inflated, but it's possible for some of these to be false-positives.
I myself am a serial song repeater. I use it a lot for focus. Noise to drown out other sounds, but familiarity so I don't start thinking about the music instead of work.
> "Several Tidal customers who have been confronted with these logs pretend to have played songs by the artists as many times as illuminated."
Is translation so much harder, or simply not a priority, or are all the claims about neural networks and language processing exaggerated?
i.e. "I never said she stole my money"
there are 7 possible meaning conveyed by stress alone, which will need to be translated differently.
also current tech has limits when translating whole pages. paragraphs introduce context and meaning that's later referenced. sentence by sentence translation is bound to lose all of that info along the way. I don't think current neural network used in mapping translations have enough complexity to keep context as they go. they might get there eventually alone as the computer scale or with the help of preprocessors.
And ... I am not sure now what. There is a online radio station thing (tidal) that has dishonest numbers.
But why would you trust those numbers outside of tidal unless you had some independnat verification ?
Is there any such independnat verification of say Apple store downloads ?
If not how do we solve both problems?
hmmm i wonder if the blockchain can be used for this
The real sticking point here is the music industry has many untouchable characters in it, witness how #METOO barely if at all can make a dent in it especially the rap music side. so unless some other artists who have real pull can show they have lost real money this will get buried. There is a story Kayne West is after TIDAL for money but how far is he willing to buck that system?
Plus the dataset is tainted - it's not trying to spot if people have been pretend listening to beyoncé songs, and picking their trees from the forest, it's that every tree in the forest has been artificially raised up. If you don't know that how AI tell?
I do agree on your second point though.