2009 - The Pull Music Paradigm Shift Emerges
January 21, 2009 
Here’s my prediction promise the project I’m working on: I intend to end the need to promote music. Outside of the social promotion people do after they discover music, artist-initiated self-promotion should vanish like the dinosaurs. But, that’s a mid-term goal. In 2009, I want everyone that thinks about the music business to have the Pull Music Paradigm Shift on his or her radar. If you have a music industry plan that begins in 2009, you have to put some of these eggs in your basket.
I’m intentionally jumping the music discovery and recommendation rails and calling it all the Pull Music Paradigm Shift because the music discovery tools that you’ve experienced so far haven’t shifted more than a bit of behavior. However, the stuff that’s about to come out of multiple R&D labs will shift an entire industry from furious push-promotion to pull-ready optimization as fast and as sure as Google changed ordinary business.
Over the last ten years, Google unwound two hundred years of the ability to price ‘business’ based upon whom you know and whom you take out to dinner. Google’s relentless dedication to organizing the world’s information has given us all dead-simple access to the knowledge of where stuff is and what it costs. Now, for just about anything consumers want, they can ‘go direct’ at the lowest price; no middlemen needed or included.
The Pull Music Paradigm Shift will change the music industry like Google changed the business world. Push middlemen with secrets and access to exposure and success will give way to a new layer of pull optimizers and pull experts that are skilled at leveraging pull technology and navigating the Shift.
I’m compelled to reiterate that this is a business plan, not a prediction. In this post, I am going to describe how it works and what will be different as pull overtakes push (promotion).
Like any technology, what I’m describing is evolving; it doesn’t and it won’t work perfectly at first, but it doesn’t have to. The world is drowning in music. The technology described here is a productivity tool that will be used to chop the haystack down from 30,000,0000 songs (plus the million songs a year that are being uploaded to the internet now) to 30,000 to 3,000 to 300 songs (needles). Moreover, it’s only going to get better over time.
The technology and methods you probably haven’t you seen yet.
I’m not talking about simple recommendation tools like Genius from iTunes, and I am not talking about any existing websites that currently feature a bit of music discovery technology. The Pull Music Paradigm Shift leverages everything you have seen today plus core technologies that will be available in 2009/2010. Here’s a rundown (1,2,3):
1) Music Experience Interfaces
Music Experience Interfaces (MEIs) will enable users to chop through mountains of music in minutes. MEIs can be visual, tactile, haptic (think Wii controller) and auditory all at the same time. Imagine using your hand to skim through music as it’s visualized on a wide screen, while rapidly listening to overlapping clips to find the exact songs or the music experiences (playlists) you desire. Have you seen the user interfaces in the movie Minority Report?
The video above is not a mockup; it’s footage from a working application (under development) created by Paul Lamere’s team at Sun Labs in Massachusetts (The AURA Project). Paul and his team have some of the smartest music discovery technology I have seen yet. If you are an executive working in the music industry, make an appointment to visit Sun when you’re in the Boston area. Contact me for additional information.
Here are a few links to people working in this space:
2) Funnel Filters
Every great music experience interface has to be attached to the ability to intuitively and rapidly funnel music.
Slicing, dicing and funneling music implies that music can be scored, ordered and sorted. The uniformed pushback I always get about filtering music is that somehow the process will force music into some sort of most-common denominator, homogenized hell. You can shelve that thought. Music can be scored, ranked, ordered and funneled against any song set you can imagine. If there’s a segment of the population that’s looking for sad country songs about love and war but not about dogs, that sort of sound like Kanye West and U2, then that’s where the user will begin his journey. Here are some of the ways songs can be scored:
- Similarity scored - how completely similar is a song to all songs in a set (any set).
- Sonic attributes scored - scored against certain sonic attributes such as melody or beat.
- Emotional content scored - the highs and lows (topology) of a song or song segment as compared to all the songs or song segments in a set (used in synch licensing).
- Market demand scored - songs are scored against hits.
- Lyric content scored - the occurrences of lyrics compared to songs in a set.
- Meta tag analysis scored - the occurrences of meta tags compared to songs in a set.
- Sounds-alike scored - does a song sound like songs in a song set.
- Market traction scored - scores based upon sales, shares, downloads and plays.
- Professionally scored - as scored/rated by music industry professionals.
- Crowd scored & social ‘friends’ scored - scores are crowd-sourced.
- Price scored (free or not) and proximity scored (geographic measurement).
Once again, the ability to score and funnel music will be as simple as adjusting faders, dragging songs and waiving your controller. The user’s ability to chop through a mountain of music will only be limited by how fast his or her ears can process overlapping clips while visually subdividing a full screen of information.
3) Distributed & Cloud Computing
While some of this technology will be accessible through the browser, there’s a lot of it that will require the user to download and install an application. To leverage the power of your computer, to avoid the added costs and hassles of licensing the major label content that you already own, and to create the ultimate end-user experience, expect applications to process music and the user interfaces locally, while shifting processed songs and metadata in and out of the cloud. This bit of the pull-technology-kit may take the longest to reach consumers. However, I expect more than a few people within the music industry to be using much of what’s described here in 2009/2010 to locate the gems buried in the beach. (I’m also looking forward to the mix of music discovery and cloud computing options that Sun will be shopping over the next year or so.)
New Business Models Needed
Music is still going to be sold and music is going to be free. However, I expect pull music technology and propositions to further sponge up the disposable time that people can invest into music consumption. As I have said before, when you attach this stuff to a database of one or two million songs, everyone is going to find the stuff (music or the general music experience) they are looking for. Lots of new music will be discovered, which will further dilute the time people spend on mass-marked (pushed) music.
So, what are the behaviors that shift as we transition from push to pull?
Prior to Google nobody talked about adwords, adsense, search engine optimization (seo), search engine marketing (sem), relevancy, organic search results and things like link farms. The Pull Music Paradigm is going to birth its’ own set of similar verbs and practices that didn’t exist yesterday; here are a few to consider (A, B, C):
A - Song Data Manicuring
To enable the creators of Music Experience Interfaces to EASILY obtain (mine for) the information needed to funnel music in a myriad of ways, song owners will be driven to be the custodians of every statistic, analysis, review, tag and chunk of relevant data they can attach to their songs.
B - Song (Data) Planning
The song scoring and funneling methods described above are also a reflection of the trend analysis and trend visualization criteria that will be at the fingertips of just about anyone. Sixty minutes in front of a flat screen full of information will be telling for any songwriter or music industry professional. As sure as you believe that featuring a kazoo in your next song probably isn’t going to work, you’re going to quickly develop a sense for what will. Will this make music creation formulaic? No never. Anomalies and outliers are still going to bust trends every day. However, marketplace knowledge will be as accessible and usable as the repertoire of music knowledge any artist brings to the songwriting table.
C - Song Data Gaming and Validation
As sure as humans work to trick Google to gain higher organic search standings, there’s going to be song data optimizers that will attempt to profit from gaming the music funnels; it will be a cat and mouse game. And, just as Google tirelessly works at achieving relevancy (you usually get relevant information from Google), the creators of Music Experience Interfaces combined with the data aggregators will work to ensure that the music funnels are also accurate and relevant.
The End of Mass-Market Music Promotion
The result of music promotion can be summed up as something called market traction. Market traction is just one-way to score (measure) and funnel music. The visualization (the graph, the plot) of music that’s mass-marketed is going to be obviously different than the visualization of music that’s found using other funnel criteria and then genuinely promoted (recommended) by identifiable segments of the population. As long as there are record labels that still generate billions of revenue from music sales, traditional music promotion will continue. However for everyone else, music promotion is something that fans will do after they filter, find and discover new songs. Since the everyone-else method will be far more cost effective (now that the product is now 99 cents or less), enduring and visually perceptible, I predict plan that mass-market promotion will have a short shelf life.
I don't use Music Think Tank to overtly push my projects, just my ideas. However for the sake of credibility, I can tell you that Music Xray is where I am working on much of what I described in this post. Don't expect miracles. Music Xray is going to take two to three years to fully build.
Reader Comments (20)
Thanks for the interesting article. It will definitely be interesting in to see where future technologies lead us.
And I also have to mention that our next song will feature a kazoo, and it's going to be AMAZING! :P
Dear Bruce,
With a few hundred feeds - mostly relating to the music industry - fighting for my attention throughout the course of any given 24 hour period, I must say, your post regarding The Pull Music Paradigm Shift caught my attention. The AURA Project graphic caught my eye so I committed to the entire post hence this response. Having spent my adult life in some facet of the music industry (and continuing to do so), I'm clearly influenced by shiny things..haha.
I'll get right to my point, I feel that the entire basis and premise for this project is utterly contradictory to what I perceive are your good intentions. Insisting that technology can replace an emotional connection to a composition is woefully short-sighted. Reoccurring instances in scoring/metadata does not mean similarity will breed a similar emotional response for the listener. I think you could agree that all great artists that are now widely replicated once defied genre.
To expand on my point about scoring, the wave of industry arguments with regards to its own demise have continuously laid blame on external factors (p2p, ability to create non-degrading copies, etc.). However, in reviewing just a sampling of the signings by larger labels (encouraged by radio) in the past decade, it's abundantly clear that many of the A&R staff had been following the very formula that ultimately led to their downfall. A formula that is strikingly similar to the methods you are proposing for future music discovery.
When you say "The world is drowning in music" what you're really implying is "most artists are insufficient and don't deserve my time because their music is not economically viable". For the sake of argument please take a moment and ask yourself, "What if economic viability played no role in future artistic endeavors?" Do you think music created under these pretenses would have a deeper impact on the listener?
Ok, admittedly, removing economic viability from the equation is difficult to imagine. However, the definition of the economic viability of music is currently undergoing dramatic shifts and will continue to do so. As such, using past industry practices as benchmarks will more than likely not work as part of the discovery method.
The music experience interface seemingly has nothing to do with music at all. How can one sort through music if they never actually hear it? This makes absolutely no sense. I watched the video but it was simply album covers swirling around. Again, it looked pretty but had zero ability to turn me on to an artist or a song I had never heard. While I will sometimes go into a record store and buy an album based solely on its cover, that is certainly not the main way by which I discover new music nor would I imagine that it is the main way others discover music.
Recommendation engines ignore the fact that discovery is about the journey to the composition so you as the consumer/listener has the ability to discern future personal likes and dislikes. Have you ever heard a songwriter say, "This is what my song is about and all other interpretations of my work are incorrect."? The very notion of the previous statement is absurd. There is no algorithm for personal taste and the more music one is exposed to, it becomes less likely that an algorithm will ever prove effective.
Contrary to your statements, self-promotion is the only form of effective promotion. No one will ever be as passionate about the music as the artist/songwriter that created it. I, along with nearly all of your readers, never would have discovered this blog if it were not self-promoted. I am ultimately writing this letter because I was moved by the content of this post. If this content did not effect me emotionally, I would not have engaged in the writing of this response. The parallels are pretty easy to draw.
It's not that the music business is failing - it's the simple fact that it should never have become the business it has become over the past 25 years. When the emotional connection was replaced by sound-alikes, that's when things really started falling apart. I'm concerned that scoring, in the way it is described in this post, may have a similar negative impact.
Again, I do believe that your intentions are good but the way in which you are trying to complete your objective should undergo greater consideration. Thanks for your time, for reading this self-promoted response and I will continue to follow the progress of this very intriguing project.
Kind regards,
Josh
"Insisting that technology can replace an emotional connection to a composition is woefully short-sighted. Reoccurring instances in scoring/metadata does not mean similarity will breed a similar emotional response for the listener. I think you could agree that all great artists that are now widely replicated once defied genre."
Well, that all depends on how one does the scoring; if you want to know how to measure an emotional response, you need only ask a behavioral scientist. I can imagine that in a few months time, people in a target demographic will pay for their music access with a few minutes a day/week under a helmet/hairnet-like device plugged into a USB port. The use of technology to streamline the path to the desired result is inevitable, IMO. If a given technology can give me a deeper and more powerful musical experience and do it more quickly, conveniently and cheaply, I say BRING IT ON, and do so without delay.
I'll buy it : this "thing" sounds amazing. a new way to discover new things. A tool you can use, alongside with your personal taste and thoughts, and choose parameters (i don't understand why pricing, crowd and music professionals could help finding the best songs we could listen to).
The old tools are dead, right ? (labels, promo etc...).
Just hope this software is not going to bug, good idea to compare it to Minority Report... anyway...
I think you are way ahead of the curve here. For the pull paradigm to really work, those MEIs would have to be ubiquitous. The pull mechanisms have to be in place first. Until then, the only way to reach those consumers who aren't yet using a MEI to discover new music is active push promotion.
I would love to switch to pull, but I think it is quite a bit too soon to make it my (only) business model.
Josh,
I wouldn't jump the gun so quickly and declare this technology useless. We need to see it first. I rather like your romantic view of great musicians ("all great artists that are now widely replicated once defied genre"), but unfortunately I don't know how accurate it is. All artists work within a genre with acoustically similar peers. I have little doubt that science can use algorithms to make highly accurate groupings of similar sounding artists. That all of the artists in a given grouping may not hit one's emotions the same is besides the point. A tool like this increases the likelihood that you will find artists that give you an enjoyable emotional experience. I don't think it homogenizes anything either. I would imagine you can make many different, unrelated searches. "I really like Shakti and want to hear other Rock/Classical Indian Fusion", or "I really want to hear band's that sound similar to Weezer."
This device is a decision tool that cuts through the clutter and strips down to the music that one really wants to hear. This sounds infinitely better than Pandora or any other recommendation tool currently out there. I say bring it on.
Hello to all.
Very interesting read.
You enthusiastic tone engages my curiosity even a bit further. You make us feel as if your concept is onto something very big, and I believe it is.
On a personal level, music discovery for me is somewhat of a complicated task. I have hardly ever encountered a band I like a lot thanks to a recommendation system. FineTune put me onto a couple of bands, so has Last.fm. TuneGlue’s Audiomap helped on some occasions, but none of these services really contributed to diversifying my musical tastes (this of course also has to do with the fact that I’m pretty elitist when it comes to music - my brain discards enormous amounts of bands I don’t consider worthy of the bands I already love for reasons I have yet to understand. This makes my take on this subject kind of biased).
Has traditional push-promotion music marketing strategies helped me discover music? Absolutely not, or maybe on some very isolated cases that I can’t even recall (I think I discovered the band ‘Fall of Troy‘ via a MySpace banner for instance. Then again it’s hardly a band I would spend money on).
It is hard to say if a model like the one Bruce presents here is going to be a new paradigm for music discovery, but I do believe his views are accurate in terms revolutionizing the current established models. If orchestrated properly, new visual and interactive platforms can only help music lovers seek what they’re looking for, and the association of data filters to songs by scoring, ordering and sorting those songs is a great idea.
This sentence helped me visualize the concept a great deal:
"[...] the ability to score and funnel music will be as simple as adjusting faders, dragging songs and waiving your controller. The user’s ability to chop through a mountain of music will only be limited by how fast his or her ears can process overlapping clips while visually subdividing a full screen of information."
Imagine an interactive platform that helps you scoop out music based on precise criterias by manipulating beautiful, high-end, user-friendly interfaces (much like taking a spoon and intentionally scooping out that portion of the ice scream carton containing an extra amount of cookie dough). No more clicks and page refreshes. No more getting lost in a maze of navigation queries.
Just music, less music, then THE music.
I defend your vision and am most curious to know what is to come of it.
(Some of the text above was taken from a blog post I just wrote on the matter in my Gigdoggy blog. Intended as a comment, I got carried away and decided to write a post instead :)
Good luck and please keep us posted on your progress!
i wonder how it would handle noise bands? i dont think you can score final or merzbow
I just discovered two band here...
thx austin :)
Experimental noise bands would probably be the easiest to to score and funnel. Extreme music breeds extreme niches
Bruce, I always relish your posts - very thought provoking, and highly respect your focused exploration of new music discovery technology - now more than ever the mountain of new music is overwhelming!
My question is (and I really am being unbiased here, as my tech knowledge is neglible): would what you are envisioning have found the historically significant/genre defining artists in POPULAR MUSIC (obviously subjective term, so please think concept more than specific artists), BEFORE they became popular, such as:
Muddy Waters, Howlin' Wolf, BB King, BEFORE there was ELECTRIC blues
Louie Armstrong, Chic Webb, Benny Goodman, Coleman Hawkins, Buddy Rich... BEFORE there was JAZZ, then SWING
Dizzy Gillespie, Charlie Parker BEFORE there was BEBOP...
Little Richard, Fats Dominoe, Elvis, BEFORE there was ROCK, then Dylan, Beatles, Stones, Grateful Dead, Hendrix, Pink Floyd...
Grandmaster Flash, Sugarhill Gang, Run DMC, BEFORE there was RAP/HIP HOP
And so on, in POPULAR MUSIC...
WE DON'T KNOW what will come next now any more than listeners did before all of the above and obviously MANY, MANY others (whether we like them personally or not).
So can technology FIND the next wave(s) of NEW artists? How can it search if the genre doesn't even exist, just as all the aforementioned genres didn't at one time?
Please believe me, i hope there is a relevant answer, and if so, i can't wait. If not, then it seems the perpetuation/homogenation question still looms...
Thanks for considering!
It is a very interesting concept and I wish you the best of luck.
It all depends on how it will be carried out but I think it has great potential in 2/3 years. Until then, I'll carry on with the push method but I'll definitively keep an eye on what you're doing because I think it is promising.
As with new concepts and ideas, it's hard to visualise how people will respond to it but I tend to agree with you that it could work.
Some thoughts:
Technology closes the distance between people. Technology may or may not close the distance between music that I like and music that I am told I may like. As a musician I don't "get" Pandora for instance.
And then technology always suggest a new paradigm is replacing the old paradigm yet as Marshall McLuhan has written "new media is always populated by old media". Take the CD for instance, at first it was populated by old media - digitized master tape recordings.
People who believe that technology is driving our interactions are missing the point - we ourselves are technological devices, invented by ancient bacterial communities as a means of genetic survival. When you take nature out of the game what is left? Humans and animals respond to the emotional root of music. Technology can never replace that.
More of these thoughts of mine can be found here - Thoughts On Social Media and Marketing
I'm really trying to imagine what criteria would have been used to choose my current music library. I don't doubt they exist; I just can't imagine the full extent of what they would have been for the proposed paradigm to create the exact same result.
As a consumer, I can't imagine this saving me any time in the long run. Perhaps it would yield more relevant choices for the same amount of time, but I'm not sure how useful that is in the big scheme of things. In fact, I can imagine my ADD-driven brain spending more time, getting deeper into it and wanting to sample more than I ever would have when using a more traditional way. With near infinite relevant choices, where do you stop? Today, if I hit a long streak of bad results, I leave it. This would be like addiction to an endless videogame.
Also, I think people are going to continue to want some push marketing in acceptable channels and venues to prompt interest in pulling specifics. Surely, marketers will take the patterns of what I've pulled and craft a targeted message to me, if not for music, then for some product related to the interests I've shown through my choices of music. As far as generating revenue to support the system, unless the operators have Consumer Reports type resolve, I suspect "payola" of some sort will be attempted.
However, by all this nitpicking, I don't mean to imply this will fail, or is the work of the devil, or any such luddite negativity. It's nothing less than fascinating, but I do suspect there'll be an unanticipated human reaction or two surprising the developers along the way.
I believe the vision Bruce presents would be an excellent improvement of existing music discovery technology. These new ideas are certainly exciting and astonishing.
However, if this technology is applied at the large scale, I think that it could do some serious damage to our already damaged quality of "POPULAR" music. It will send us further down the road of music replication, making artists EVEN MORE concerned with turning a profit than sparking emotion. Bands today already feel the need to pick a set of genres that their music "fits" into, which obviously wasnt the case back when when these genres were being formed! People wrote music that impacted them emotionally, and did not care about the title their music received, or how "experimental" it might sound. That is why the greats mentioned in above posts have stood the test of time.
I believe this technology, applied at the large scale will further reduce the diversity of popular music, and popular music will indeed become formulaic.
If applied at a smaller scale, it can indeed be an excellent tool to help guide music lovers to new music.
I do not know if any of this made sense, but I just felt the need to ramble. haha. And thank you Bruce for this extremely interesting and insightful post.
Thank you for the thought-provoking piece. I take that in practice the Music Explaura would be navigated more slowly, would allow users to listen to the music as they passed by, and would be more multidimensional.
A few comments on some of the ways you suggest that songs could be scored:
Some technology companies are already offering this service to songwriters and record companies. A possible downside is that groundbreaking but low-scoring demo pieces may be modified conservatively by songwriters in order to obtain a higher score, or may never make it to the recording studio. The latter is also possible for songs with extraordinary lyrics.
What about the quality and meaning of the lyrics? Artificial intelligence will not be able to do lyrics justice - certainly not for the current generation of artists.
These last three can all be heavily influenced by self- or mass-market promotion. I don't see how you are going to do away with promotion.
Finally, even if people predominantly use such tools to actively find music, people will still be reachable by mass-market advertising or promotional techniques. What do you think?
Anyway, the aim of making it easier for artists to find a sizeable audience without needing to self promote or be promoted is a good one.
Full disclosure: I have a bias in favour of those who are shy and write brilliant lyrics (such as Andreja Volenec).
Thanks to everyone that responded to this post. I usually respond to comments immediately, but this time I came down with the flu within 12 hours of publishing this post. I carved off some questions below. Your questions are in italics and my responses are below each question.
Josh Preston
Insisting that technology can replace an emotional connection to a composition is woefully short-sighted.
Josh - I don’t believe that I insisted that technology can replace an emotional connection to a composition. I have to ask - did you really read the post? People will use tools like this to find songs; they will have emotional responses; they will promote songs to their friends; the systems will recognize the activity; more people will find the songs (based upon activity - because that is how SOME people will choose to filter - i.e.: let other people do the discovering and early adopting); more people will then have more emotional responses; then even more people will promote the songs to friends; the systems will recognize the additional activity; and so on...
Jim Offerman
For the pull paradigm to really work, those MEIs would have to be ubiquitous.
Jim - I think ONE person could use an MEI effectively. Wait until you find out what the small (shrunken) A&R staffs at some labels are (and will be) using…
DG
So can technology FIND the next wave(s) of NEW artists? How can it search if the genre doesn't even exist, just as all the aforementioned genres didn't at one time?
DG - Absolutely. I will bet my last dollar that the tools will enable those that are looking for the anomalies and the outliers to find them. Furthermore, the same tools will enable these searchers or seekers to rapidly playlist and share the gems that they find. And, as I said to Josh above - it all becomes a reinforcing loop that continues out to the extent that the outliers and anomalies resonate (generate an emotional response) with the early adopters (fans) these songs reach. Think of it as concentric circles or waves of enthusiasm that are captured (as statistics) and measured (against other waves) that are all visible for anyone to sample (hey what’s going on over in that part of the music universe). Like it has ALWAYS been, the waves (the circles) are only going to grow if the songs resonate.
Dave Allen
Humans and animals respond to the emotional root of music. Technology can never replace that.
Dave - OK. Technology can and does short cut the path from creator to emotional response within the recipient. We could go back to cave drawings and tribal drum beats, and then wait until a passerby pops wood over what he sees or hears…
Chris
If this technology is applied at the large scale, I think that it could do some serious damage to our already damaged quality of "POPULAR" music. It will send us further down the road of music replication, making artists EVEN MORE concerned with turning a profit than sparking emotion.
Chris - No doubt - damage will be done to popular music. I believe this technology will cause so much observable fragmentation that the bracket of popular music will be hard to define. At first, everyone may try to ‘time the market’. However in the end, I bet most will go back to just making the music they love, as dedication to being true to the music that resonates with one’s self will have more observable impact than contrived attempts to be popular.
Frank
A possible downside is that groundbreaking but low-scoring demo pieces may be modified conservatively by songwriters in order to obtain a higher score.
See what I just wrote in response to Chris’s comment above.
Artificial intelligence will not be able to do lyrics justice - certainly not for the current generation of artists.
I believe you are thinking that we would rely on computers to extract lyrics. That’s not the case. Lyrics will come to us as - lyrics (text files). Yes, lyrics can be scored effectively.
Market Traction, Professionally Scored and Crowd or friends scored can all be heavily influenced by self- or mass-market promotion. I don't see how you are going to do away with promotion?
As I said, mass-market promotion will be observable. Genuine promotion that is being done by crowds that are having true emotional responses (as in sharing, playlisting, etc) will also be observable. Also (as I also mentioned) mass-market promotion will be done as long as people can make money doing it. However, for almost everyone else (and this number is growing rapidly), mass-market promotion has no ROI when the product is almost FREE. I expect smart people will figure out how to let the song + the crowd’s emotional response…do the promotion work for them.
One more thing - I don’t expect any kind of promotion to change the professional opinion (score) that a true professional may or may not have about a song…
Hi Bruce,
Thank you very much for your latest post. I hope you're over your flu! A few comments and clarifications:
When I said that Artificial intelligence will not be able to do lyrics justice for the foreseeable future, I didn't think that identifying the lyrics was the problem. I meant that I don't think it will be be possible to automatically analyse the quality and meaning of lyrics effectively for users. It may be possible to automatically identify certain styles or to assess the complexity of lyrics. It may also be possible to make some other guesses about the lyrics. But by analysing the lyrics alone, AI won't anytime soon be able to answer questions like: "Are they any good?" "Will they move me?" "Are they profound?" "Are they memorable?"
The idea that genuine promotion will be observable is interesting but is it true? Suppose it were true, I suspect that it would still be the case that crowds would still predominantly promote from the subset of artists that had been brought to wide attention through the efforts of the artist or their label.
There is a fascinating Pulitzer prize-winning article in the Washington Post on Joshua Bell's attempt at busking. Mark Leithauser, a senior curator at the National Gallery in DC, believes the explanation is context. He thinks that if a $5m painting was hung in a restaurant with a $150 price tag, no-one would no notice it.
I think the relevance here is that the music of a successful well-known artist will be judged by most people to be better than the same music would be if the artist were unknown. People use signals (context being one) to interpret the world. Promoted songs will be more favourably viewed. I suppose the hope would be that directed promotion would become less cost-effective in this new environment and decline anyway.
Professionals may be less susceptible but I think would still be unconsciously influenced by external signals. The additional problem is that pros will tend to listen to and review the most popular and high-profile work even if they don't end up liking it. Initially obscure work will find it hard to get reviewed and may remain obscure.
So the challenge is big, but I hope I'm proved wrong and your goal is more tractable than I think it is. Good luck!
Frank
Frank - thanks for another thoughtful response.
Here's an interesting white paper (PDF) on using lyrics to predict hits.
http://www.hpl.hp.com/techreports/2005/HPL-2005-149.pdf
Can the same methods be used to compare (score) one song against any group of songs? It would seem so?
It seems like the visualization of mass-market exposure would look like an atomic bomb compared to the firecrackers that small crowds would resemble. We will see... But, maybe nobody will care?
Don't really buy the analogy (paintings versus songs). I buy the art that I like and I attempt to simultaneously purchase art that will rise in value (and I am not very good at it). When it comes to music - I just buy what I like. Moreover, songs are essentially free. However, I get what you were trying to say.
Yes, kids need signals and the songs kids download become the sponges that absorb the memories of their youth. However, beyond 22 to 23 years old, for the remaining 50 to 75 years of one's life - I believe (however, I have no clickable proof) that signals are far less important to music consumers. (Last night my teenage daughter put three hundred songs onto my Mac for me to try. I am starting to wonder where (if any) the signals came from?)
You said "promoted songs will be more favourably viewed". Ouch.
I suspect somewhere between the hype (my forward-looking post perhaps) and what's possible - is a product that professionals (and some day consumers) will use to rapidly filter the thirty million songs in existence + the million songs a year that are being uploaded now - to find songs that were never promoted. These found songs could be then:
- promoted
- tied to a brand - and then promoted
- shared between groups of friends (crowd promoted)
- placed into soundtracks
- placed into commercials
- used to find the artists that someone (record label?) may want to promote..
- etc..
If the goal is to shortcut the need to self-promote - prior to being promoted (any form at all) - the PRODUCTIVITY tools that enable the promoters (professional or consumer) to discover new songs and artists (without signals and from within the ocean of possibilities) should expose the urgency to aggressively (underline) self-promote as...somewhat pointless. I mean, how can an artist justify the negligible ROI when he/she starts to frequently hear the stories about the artist that was scooped out of his basement by way of the song he licensed into a major motion picture - no promotion required. When this (discovery without promotion) starts happening regularly, it will be clear that more songs and less promotion are the tickets that will get you found in the 'magic discovery box'...
Hi Bruce,
Thanks for the interesting reply. Thanks also for the white paper link. I checked it out.
It references another hp paper which deals with lyrics analysis. However, it appears to me that the analysis just looks at the frequency and synonyms of individual words rather than the meaning of phrases, let alone verses or entire songs. An automated analysis could perhaps help to spot manufactured hits but would have difficulty with originality, brilliance, subtlety or humour. Such a program may not distinguish between:
"Please don't leave me. Love me." and
"Please don't love me. Leave me."
Returning to the art analogy, could an automatic analysis identify beauty? That's an issue with tone as well as lyrics (and melody, etc.). I'm not sure if current music analyses look at anything other than melody, tempo and rhythm. Some singers have beautiful voices, e.g..
Your proposal leaves scope for human crowd and pro recommendations, which can work for lyrics (and tone). So it's not an obstacle to your system being useful if automatic lyric analysis is absent or is limited in scope for now.
I would be concerned that if you included some kind of black box lyric analysis, the software might be perceived to be more intelligent than it actually is. I did some research in AI in a past life and, although I haven't kept up with the field, I just doubt that progress will be quick in the area of language comprehension. I believe musical analysis is much more practical.
However, it may be possible to make some progress with lyrics without real comprehension. If automated analysis reveals that people who like A tend to like B, C and D, then people familiar with B, C or D might guess that the A’s lyrics are of a similar standard and sensibility. If B, C and D had lyrics scores somehow, the computer might even guess a score for A. It would still be a guess though… and a crude measure.
By the way, I love the end of your latest comment above.
Frank
Nice article Bruce, where future technologies lead us interesting. Thank you for sharing with us. For Songs Lyrics Visit