Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Programmers generate every possible melody in MIDI to prevent lawsuits

musictech.net

Post reply on HN
Print view View original

41–50 of 340 posts

Re: Programmers generate every possible melody in MIDI to prevent lawsuits

#41

Post by Muller20 »

throwqwerty wrote: post #33
drno123 wrote: post #31

Earlier quoted context omitted.

Yes, I meant to run 2D CNN over generated spectrograms. We do something similar to classify some specific emissions in RF with good success. As for scoring/classification, you can start by having an output of CNN say whether the song is catchy or not.

why run a CNN over a spectrogram? besides what I said (convolutions in frequency domain are multiplications in time domain) an FT is linear. if classifying using those features were effective then your CNN would've learned the DFT matrix weights from the original signal.

because convolutions are faster during training and after the deployment and they are also easier to train.

Re: Programmers generate every possible melody in MIDI to prevent lawsuits

#42

Post by tunesmith »

It's about fourteen kinds of ridiculous, as summarized in other threads. No rhythms, no meter, no tempo, melodies are longer than 12 notes, it's diatonic, single octave, no concept of underlying harmony, the headline is literally false, etc.

Some of the copyright lawsuits are dumb and this is effective satire or performance art but that's all it is.

Re: Programmers generate every possible melody in MIDI to prevent lawsuits

#43

Post by kd5bjo »

throwqwerty wrote: post #33
drno123 wrote: post #31

Earlier quoted context omitted.

Yes, I meant to run 2D CNN over generated spectrograms. We do something similar to classify some specific emissions in RF with good success. As for scoring/classification, you can start by having an output of CNN say whether the song is catchy or not.

why run a CNN over a spectrogram? besides what I said (convolutions in frequency domain are multiplications in time domain) an FT is linear. if classifying using those features were effective then your CNN would've learned the DFT matrix weights from the original signal.

A spectrogram has time on one axis and frequency on the other, so the ultimate result is a multiplication in one dimension and a convolution in the other. It can be used to show things like when a note starts and stops in a piece of music, which is difficult in either purely-time or purely-frequency space.

Also, it’s computationally intractable to individually train 2^N weights. What a CNN does instead is train a convolution kernel which is passed over the whole domain to produce the input for the next layer; by operating in frequency space, it’s considering the basis functions e^{j omega +- epsilon} instead of delta(x +- epsilon)

Re: Programmers generate every possible melody in MIDI to prevent lawsuits

#44

Post by itronitron »

Leszek wrote: post #17
Double_a_92 wrote: post #16

Earlier quoted context omitted.

A tl;dr would be nice it you post such a long article that doesn't directly seem to have anything to do with the current topic.

It's actually very relevant, the TL;DR (from memory) is that lawyers (and judges) aren't programmers, and the source of data (referred to as the "colour of the bits") is as important as the actual bits of the data. Thus "I have this bit pattern as part of a bit enumeration" is different, to the legal system, than "I have this bit pattern because I specifically created it".

A counterpoint to that is that genetically modified organisms can be patented and those are effectively random perturbations of current 'state of the art' which are then selected for fitness.

Re: Programmers generate every possible melody in MIDI to prevent lawsuits

#46

Post by Lewton »

lillesvin wrote: post #18

Adam Neely (professional musician and YouTuber mostly about music theory) did a video about this with the guys behind it: https://youtu.be/sfXn_ecH5Rw

This really needs to be at the top, as it answers a lot of the criticisms found here on HN

Re: Programmers generate every possible melody in MIDI to prevent lawsuits

#47

Post by tzmudzin »

Robin_Message wrote: post #40

I am not a lawyer, but as I understand copyright law there seem to be to be two problems with this: - a copyrightable work must include some human creativity. It seems to me that an enumeration of possibilities might be creative, but there is no way an individual element of that enumeration can be considered creative. - Copyright depends on copying . If you release a song with a catchy melody stolen from another song…

> What matters is which one you copied, not the existence of alternate copies.

Not a lawyer either, but "I came up with this myself" is the standard defense of the artists, with mixed success.

Re: Programmers generate every possible melody in MIDI to prevent lawsuits

#49

Post by hobofan »

albertzeyer wrote: post #36

The code is written in Rust. The core algorithm can be found here: https://github.com/allthemusicllc/atm-cli/blob/master/src/ut... (Specifically the function gen_sequences. It basically uses multi_cartesian_product to iterate over all permutations.) Isn't it a bit of wasted space to store the generated data on archive.org ( https://archive.org/download/allthemusicllc-datasets )? I mean, as we see, it's trivial to cre…

I'm not sure if it would hold up legally anyway, but with it already being generated I can see it working out a bit better.

If you only provide the code as a means of generating any melody, you could just as well replace it with any music instrument and claim you can generate any melody with it.

Re: Programmers generate every possible melody in MIDI to prevent lawsuits

#50

Post by willvarfar »

albertzeyer wrote: post #36

The code is written in Rust. The core algorithm can be found here: https://github.com/allthemusicllc/atm-cli/blob/master/src/ut... (Specifically the function gen_sequences. It basically uses multi_cartesian_product to iterate over all permutations.) Isn't it a bit of wasted space to store the generated data on archive.org ( https://archive.org/download/allthemusicllc-datasets )? I mean, as we see, it's trivial to cre…

At the very beginning of this excellent interview https://www.youtube.com/watch?v=sfXn_ecH5Rw that I found in the comments here, they explain this.

Apparently, for this to work, they have to "affix them to a physical medium".

Strongly recommend watching that video! Excellent info.

Post reply on HN