Tech: Technology predicts/composes hit songs

Subject: Tech: Technology predicts/composes hit songs
From: Helen
Date: 14 Nov 06 - 05:16 PM

Hi all,

I'm pasting in the entire article because it won't stay on the website for long and I think this is a long term discussion topic. For me, it is about record companies trying to turn a quick buck, and potentially - or already - throttling the life out of real musical creativity by forcing all commercial music into a narrowly defined hole. This article is about the development of technology which could be used to make the problem even bigger, depending on whether and/or how it is used.

There are a lot of issues involved in this, and some are touched on in the article. On the other hand, the concept of the technology is pretty clever.

Helen


Pop perfection or manufactured conformity? http://www.abc.net.au/news/arts/articulate/

Posted by Adele O'Hare: Norah Jones: A lot like Van Halen?

Is musical taste a uniquely human, personal thing, or can a computer program mimic it well enough to pick hits and even create them? The Guardian reports that a handful of companies are "using psychological research and mathematics to try to create the perfect pop song".

Earlier this year, Articulate looked at a program called Echo Nest that analyses the pitch, key, rhythm and structure of songs, and surveys people's opinions of them to predict what will be a hit.

The new technology goes even further. A company called Platinum Blue Music Intelligence stores data on every song that has ever made it into the UK Top 40 or the US Billboard Hot 100 since the 1960s – plus plenty of songs that flopped.

Using "spectral deconvolution", the company's software can pick out 40 pieces of data about the deep structure of each song – including its melody, chord progressions, instruments and "fullness of sound" - and plot the data visually, in a three-dimensional "music universe" with each hit represented by a point of light.

When plotted like this, about 80 per cent of the songs that have ever been hits are positioned together in 50 clusters of stars, with black space in between – meaning any popular song is likely to be based on one of a relatively small number of basic structures. And Norah Jones's jazz belongs to the same cluster as the pop of Vanessa Carlton and the metal of Van Halen.

Could the secret of a hit song be this obscure and mathematical? There's huge potential for record companies to use this kind of technology to predict hits and alter songs to improve their chances of chart success – but The Guardian reports that they rarely own up to it.

"It's evil," is the verdict of British pop songwriter Guy Chambers, who's written hits for Robbie Williams.

    "It reminds me of Hal, the computer in 2001: A Space Odyssey, and Hal's not a friendly computer… It's a really desperate situation. It's one of the reasons the music industry is on its arse. There are lots of reasons it's on its arse, but one is that there aren't a lot of people willing to put gut instinct first."


And what about the moral question of whether reliance on the technology will enforce conformity and crush musical innovation? Philosopher, computer programmer and musician Jaron Lanier argues that the software would fail to recognise the value of a truly original work of musical genius:

    "The problem with replacing creativity with analysis is that while, almost by definition, it will work some of the time, it creates a meaningless sensibility. This isn't just a moral point, it's the moral point. If our purpose is to please ourselves in the most average way possible, without caring what anything means, we might as well just kill ourselves. We've lost the moral authority to want anything."


Then again, there's a long tradition of artists shunning the corporate record labels and mainstream sound. If the use of hit prediction software becomes widespread among record labels and radio stations, perhaps we can expect a resurgence of music that's truly indie in response.

Subject: RE: Tech: Technology predicts/composes hit songs
From: Joe_F
Date: 14 Nov 06 - 08:59 PM

Here is the speculation of a person so marginal as perhaps to be crazy:    The popularity of a tune, at any rate since (say) 1940, has had very little to do with the tune itself. Very large numbers of customers are too cowardly to follow their own tastes; they are concerned first of all to agree with what their peers are buying. Such people may not be a majority, but they are the most reliable & profitable customers (as drunks are the most reliable & profitable customers of bars), so it is they to whom the market adapts itself. Popular songs form clusters in quality spaces because a tune that resembles tunes that have already been popular seems like a smaller risk to the manufacturer and to the customer, both of whom would prefer to be participating in a fad -- the first for profit, the second for comfort.

Subject: RE: Tech: Technology predicts/composes hit songs
From: Scrump
Date: 15 Nov 06 - 08:32 AM

I won't lose any sleep over this. I predict it will be a failure. Although the charts often seem to be full of the same old repetitive stuff with vacuous lyrics, performed by people who can't sing or play their instruments, there is always room for originality and high quality, which surprises everyone now and again. I can't see that changing.

Subject: RE: Tech: Technology predicts/composes hit songs
From: leeneia
Date: 15 Nov 06 - 09:05 AM

Well said, Joe and Scrump.

One thing the computer programmers didn't plan for is the human desire for novelty and innovation. How do you tell a computer to do something it's never heard of before? It can't, but people can.

As an example: I don't know how many times I have seen people comment on the Mudcat that with popular music we hear the same 25-30 tunes over and over. One result of that sad state is the growth of independent music releases as musicians struggle to play what they want to play.

Subject: RE: Tech: Technology predicts/composes hit songs
From: Darowyn
Date: 15 Nov 06 - 09:28 AM

I have a book here- downloadable from the internet- by Philip Dorrell. It's called "What is Music? Solving a Scientific Mystery".
It goes into mathematical matrices and Neural maps, but it has two very simple and plain tests for things like this.
1. The Grandchildren test.
This test goes like this; if music or knowledge of music is fundamentally important to humanity, it must have a beneficial evolutionary effect. Therefore any valid theory of music must be able to explain how music helps someone to have more grandchildren.
How many grandchildren have the inventors got?
2. The Luxury Yacht test.
This one goes like this. If a person claims to understand the mechanism for creating successful music, they must be able to prove this by creating music successfully. Since it is commercial success that is the issue here, they must therefore be commercially successful. So the basic question is, do the inventors have luxury yachts?
The second question is, if the device really did create commercial success, would they be telling us about it or cruising in the Seychelles?

Logically this must be nonsense, as well as for all the other reasons given above in previous posts.

I recommend the book if you are interested in going much deeper into what music is, from the point of view of a mathematician.
Cheers
Dave

Subject: RE: Tech: Technology predicts/composes hit songs
From: The Fooles Troupe
Date: 16 Nov 06 - 02:11 AM

I've got a great tip - horse 3 race 5....

Subject: RE: Tech: Technology predicts/composes hit songs
From: Arkie
Date: 16 Nov 06 - 09:13 PM

In the 30s, 40s, 50s, 60s, and I am not sure when the change came but in these years hits were determined by what people wanted to hear though tastes changed with each new era. In the last decade and possibly sooner, hits have been determined by what publishers chose to be hits.   Radio programming on the more commercial stations with larger listener base is no longer determined by what people prefer. Witness the case of the "Oh Brother" soundtrack and others that achieved sales and awards without the benefit of radio. While it happens, it does not happen often.

Subject: RE: Tech: Technology predicts/composes hit songs
From: Charley Noble
Date: 16 Nov 06 - 10:00 PM

This kind of software which analyzes decades of hit songs, their lyrics, tunes and rhythns deserves a better name. Why don't we call it DECOMPOSITION! I bet it would sell like cow paddies, hot cakes I meant to say. I will respectfully decline the royalties, which could distract me from creating my next hit song.

Cheerily,
Charley Noble

Subject: RE: Tech: Technology predicts/composes hit songs
From: number 6
Date: 16 Nov 06 - 10:08 PM

The Brill building goes digital.

biLL

Subject: RE: Tech: Technology predicts/composes hit songs
From: Richard Bridge
Date: 17 Nov 06 - 03:15 AM

I have read the occasional article on this elsewhere, and I would appreciate any views on how it is possible to represent 40 different types of data three-dimensionally. With only three axes the lowest possible number of variables per axis is 13,13,and 14, and the coincidence of data within each of those sets of variables must depend on arbitrary choices of axis scales.

Mustn't it?