‘1984’, New Rembrandt And Spotify’s ‘Fake Artists’

Life appears to be imitating art as a number of strange parallels between the dystopian George Orwell novel 1984 have begun to manifest in the form of fake art and fake artists designed to entertain the masses.

_________________________________

Guest post by Laura Kobylecky from Music Technology Policy

“It was only an ‘opeless fancy.
It passed like an Ipril dye,
But a look an’ a word an’ the dreams they stirred!
They ‘ave
stolen my ‘eart awye!
The tune had been haunting London for weeks past. It was one of countless similar songs published for the benefit of the proles by a sub-section of the Music Department.”

– From 1984 by George Orwell

In the dystopian world of George Orwell’s 1984, there is a machine called a “versificator.” The versificator makes what might be called “fake” music—songs that are “composed without any human intervention whatever.” In April of 2016, “A New Rembrandt” was revealed (1). The painting, like the songs of a versificator, was made by machines. In August of 2016, Music Business Worldwide (2) accused Spotify of “creating fake artists.” What is a fake artist? Can music be fake?

The world of 1984 is a grim place. Members of the “Party” have access to resources based on their rank. The rest of society are called “Proles.” The term is short for the “proletarian” and refers to the working class. The Proles make up the majority of society, and so the Party provides them with various sources of entertainment to keep them from getting too restless.

The versificator is one of the entertaining distractions made by the Party. A versificator generates songs that are “composed without any human intervention whatever.” The results range from insipid love songs like “Hopeless Fancy,” to the “savage, barking rhythm” of the “Hate Song”—designed to stir rage against political enemies. The novel’s protagonist describes one of these songs as “dreadful rubbish.”

But the Proles like it fine. The song “Hopeless Fancy,” takes hold among them and “haunts” London for weeks. In this case, the art of the machine seems adequate for consumption.

The versificator is an element of fiction. However, “The Next Rembrandt” is real. (3) Microsoft, a participant of this project, describes it as “blurring the boundaries between art and technology” and states that the “project is intended to fuel the conversation about the relationship between art and algorithms, between data and human design.” The project used technology to make a painting that supposedly recreates Rembrandt’s style.

The portrait is a man. He has a black hat, tilted slightly. A goatee graces his face as a white ruffled collar draws the eye. His lips are slightly parted in the gesture of a half-spoken word and his eyes are inquisitive and bright. He has never lived. He is purely the manifestation of technological innovation and he exists as “a result of analysing data from Rembrandt’s body of work.”

The portrait took over 18 months. The project was based on access and study of primary data. All 346 of Rembrandt’s paintings were collected in “high resolution 3D scans and digital files.” With the art digitized, it could be studied. An algorithm searched for data points. From the data, various facial points were determined. The result was brought into the world with a 3D printer that recreated the layers and textures of paint on paper. The final product is a pleasant “painting.” It’s nice to look at, but it won’t fool the experts—yet. In the future who knows. Is it fake art? That depends on perspective.

What makes a “fake artist”? In August, 2016, Music Business Worldwide (2) accused Spotify of creating “fake artists.” However, this isn’t quite what it sounds like. Unlike “The Next Rembrandt,” these tracks are not being made by a machine—according to the post. The article states that Spotify has been “paying producers to create tracks within specific musical guidelines.” These producers get a flat fee.

So why does this bother anyone? It’s partially because of playlists. Spotify’s playlists are big money for some musicians. A “chill vibes” piano play list can be a great way to get plays for a composer. However, if Spotify chooses to drop “their records in the playlists in favor of its own masters,” that seriously shifts the balance of power and profit. Those same composers might have a reduced chance of profit and success on this platform.

There is another problem with the, hypothetical, “fake artist.” In a second article, (4) Music Business Worldwide addresses the issue of how these “fake artists” could be driving down the “per-stream income for everyone, while lowering the negotiating power of the labels/publishers/collecting societies.” The following chart illustrates that issue:

Royalty Allocation Ratio

The problem depends on the “allocation ratio,” or how people are getting paid. The bigger the “total plays” the smaller the “per play rate”. If the total pool of monthly revenue available for royalty payments is divided equally over the total number of plays, that determines the “per play” rate for that month. Each artist or songwriter would get paid for each of their plays based on that rate. (There may be complexities like minimum payments and country variations depending on negotiation power, but the basic math is pretty consistent.)

However, it doesn’t quite set right if somebody is watering down the “total plays” by including the “flat fee” folks. Flat fee artists already got their money and they aren’t getting more from the royalty pool no matter how many plays they got. Including their plays in the pool would serve no function other than to reduce the rate that the royalty artist gets per play.

It’s kind of like if you work at a restaurant and tips are your main income. You might have agreed to take a tip share. Everybody’s tips are added up and divided equally. So you make a salary, say, $2.13 an hour and expect to work for the rest in tips. But what if a co-worker has decided they’d rather not gamble on the tips. They agreed to take a flat hourly rate of maybe $11.25. Well if at the end of the day there are 6 waiters getting hourly plus tips and 4 hourlies, but the tip share is divided over ten people, things are strange. The six waiters are only getting a tenth of the total tips, but the extra money being held back from the tip share won’t be given to the hourlies. The extra money just goes to somebody else’s pockets.

Spotify has disagreed with this assertion. The Guardian quotes Spotify:

“[It’s] categorically untrue, full stop. We pay royalties – sound and publishing – for all tracks on Spotify, and for everything we playlist. We do not own rights, we’re not a label, all our music is licensed from rightsholders and we pay them – we don’t pay ourselves. We do not own this content – we license it and pay royalties just like we do on every other track….”

However, the Guardian indicates that even if the royalties are paid they might still be “much more favourable to the company than its standard deals with record labels.” A small change of fraction of a percent of Spotify’s reported $2.8 billion total royalty payout could add up to a great deal of saved money over time. Some of the same math of the above chart could still apply.

But in the end, these are still genuine human artists making the tracks, according to reports. The “fake artists” might be people with different names, but they are people nonetheless. Nobody has invented a versificator re-creating the “Next Rembrandt” of music. But could it happen?

The Next Rembrandt was based on a collection and study of data. Algorithms were used to apply the data points. So if the Rembrandt recreation took high res scans of all the Rembrandt works, then the versificator also would require a massive collection of data to work with, as well as the tools to use that data.

Spotify has made many recent acquisitions of companies that interact with data. Forbes addresses ( 5 ) some of these recent acquisitions. One of these startups that study data is Preact. The company is described “Learning everything possible about what makes subscribers happy, what they don’t like, what they’re talking about online.” Another acquisition is “Soundwave,” a startup that tracks “what songs people played on their phones and where.”

Another major purchase on the data front is The Echo Nest (6). This company “uses data analysis and machine listening to power song recommendation, audio fingerprinting and audio analysis.” Fast Company goes into further detail about The Echo Nest ( 7 ) . They describe the way that it:

“…devours data about the music, on both the “acoustic side”–tempo, key, etc. (Echo Nest’s system crunches that sort of data in about 10 seconds for a song)–and the “cultural side”–what reviewers are saying about the music for instance.”

So it seems like the Echo Nest has some capacity to study a song, the content of it, and also to study how those particular content elements affected people (the cultural side). This data is valuable. What could be done with it?

The New Rembrandt was a study of finding data in the work of Rembrandt, followed by an analysis of that data, and completed with an application of that data. It took massive amounts of data, and tools to study that data to make a genuine “fake” machine-made painting. If someone wanted to make a versificator, capable of producing genuine “fake artist,” how would they do it? Well, it might take access to a massive pool of music, tools capable of studying that data, some motivation and the financial backing to make it all come together.

This is where the science fiction side comes back. What would motivate someone to make a versificator, one that could produce Prole-pleasing content? First, it must be examined how such a machine would affect the industry. What would a versificator do? If a machine could simulate the art of humans, with any degree of success, it would certainly shift some power in the music and tech industry. The mechanization could reduce individual bargaining power. Even the production of “filler music” could leverage negotiating power enough that major shifts occur. Who would be motivated to make this happen?

In the end, does it really matter? It’s still art maybe, just made by a machine. Does human-content have less value simply because humans didn’t make?

In 1984, the narrator watches a “Prole” prole hanging her laundry on a line. She hums “Hopeless Fancy,” a versificator song, but “the woman sang so tunefully as to turn the dreadful rubbish into an almost pleasant sound.” The song is appreciated, and it connects to this human enough for them to sing along. Is that connection enough, to elevate the product to art?

Would you listen to songs from the versificator?

REFERENCES

Guest Post on 08/10/2017 in Music Business | Permalink | Comments (0)

Share on: