Monday, January 14, 2002 2:42:57 PM
THE FUTURE OF AUDIO ENTERTAINMENT
MIT Media Lab
September 5, 2000
The Nature of Push and Pull Delivery
The distribution of new media to customers has traditionally exploited two modes of delivery. The first mode is the broadcasting of content (such as radio or
television programs) to many users simultaneously. In this scenario, the end-user has little or no direct control over what content is being broadcast at any particular time (the only feedback being via ratings measurements of the audience for a particular program or station). The second
delivery mode is where the customer directly procures the content he or she desires, as with the purchase of a recording. Both modes of delivery have since embraced the world of digital delivery, where they have been coined push and
pull, respectively.
These two means of delivering content exist because we as consumers have a need for both. At times an individual will
know exactly what he or she wants to listen to, and pull is the best means of satisfying this consumer's needs. If we enjoy a particular song to the extent that we wish
to listen to it multiple times on multiple occasions, it makes sense to purchase that recording and simply play it whenever we wish. But there are also times when we
have no strong opinion about what we'd like to hear or only a general preference for a genre of music: perhaps some pop
music, maybe some jazz, or just something new. The broadcast (push) medium exists to fulfill this need. People will always have both these basic desires, and there will be occasion for content delivery in both modes, i.e.
there will always be a need for both push and pull.
The Convergence of Services Based on Push and Pull
Technological advances in digital media delivery are gradually moving push and pull services towards a common
point. As services built on the push mode become more mature, they start to incorporate features of the pull mode, and vice versa. We are witnessing the development
of a continuum that spans their difference, and the hitherto distinct services are beginning to populate points all along this spectrum. We expect future technology will enable a gravitating of services towards the mid-point of this spectrum, where features of both push and pull will
be equally important (Figure 1).
Figure 1: The continuum of push-pull services
Convergence
Unlimited sharing (Napster, Gnutella)
Basic user-control (ClickRadio TM )
"Micro-genres" (Internet radio)
Limited sharing (mp3)
CD purchase
Broadcast
Push oriented services
Push technology over the Internet began with the introduction of audio streaming, popularized by RealAudio.
While initially an attempt at content delivery on demand, the bandwidth limitations of the early Internet resulted in
low sound quality. However, its usefulness was quickly realized in applications such as speech (talk and sports radio), and thus streaming content delivery naturally adopted the traditional broadcast model. Radio stations would broadcast news and sporting events, and these broadcasts were not limited geographically, as they are with a traditional broadcast radio station. This was the primary force behind the early adoption of streaming: access to real-time out-of- market content.
As bandwidth and sound quality improved, music broadcasts became more prevalent on the Internet. Music stations began simulcasting their signal to the Internet, primarily so that out-of- market consumers could still listen to
their favorite stations. The most successful of these were highly specialized local stations, satisfying musical tastes not catered to outside of the local broadcast area. Now, this "micro-genrefication" has taken over Internet music broadcasting, with highly specialized stations, e.g. only 1983 British Pop, or all Charlie Parker. And while radio station simulcasts contain advertisements, many of the smaller Internet-only audio broadcasts do not. This provides an additional level of user control, not available over the traditional broadcast (push) medium, but moves the application towards the center of the push-pull continuum.
The next generation of push services introduce an even greater level of interactivity and control to the end user.
Services such as ClickRadioTM allow the user to skip a song they don't want to hear. These services are currently free
and are supported with advertising. And with greater interactivity comes more information about the user, which leads to better user profiling and the ability to finely target particular consumers.
Pull oriented services
The spread of pull technology for digital music delivery was catalyzed by the development of the mp3 standard for
audio compression. As soon as computers became powerful enough and storage space became cheap enough to make
playback and storage of mp3 files realistic, mp3 files became the standard method for aggregating digital music
content. Initially users would convert, or "rip", their own CDs into mp3 files on their computers. Even before sharing of files became widespread, many people converted their music libraries simply for the added convenience. This use of digital content represents the extreme end of user control. The music selection is entirely determined by the consumer's decisions regarding which music to rip.
Of course, the use of mp3 files exploded with the introduction of easy-to-use file-sharing software, most notably Napster. Prior to Napster, it was difficult to find and share popular music on the Internet. Large private collections of music existed, but they would appear and disappear in order to avoid being discovered and prosecuted by copyright enforcement agencies. With Napster, and other peer-to-peer networking systems, each user who logs into the system has the ability to share their entire collection. And as soon as they log out of the system, their collection disappears. Since each user only appears for a limited amount of time, it is difficult to ascertain which individuals are guilty of copyright infringement. However, since many (some fraction of over 20 million total) users are logged in at any particular time,
virtually any piece of music you desire can be found. It is a music lover's dream and a copyright holder's nightmare.
Napster, and its peer-to-peer brethren (Gnutella, Scour, and Freenet) have their bases in pull driven technologies, since the end user must specify which song he or she wishes to search for. However, these technologies represent a slightly different point in the push-pull continuum. As one searches on Napster, he or she is exposed to "serendipitous hits", or unintentional results of their search that may appeal to them. It may be a different recording or mix of the song they were looking for or an unfamiliar work by the same artist. Nevertheless, the user is exposed to additional music choices, that are presented to them. Additionally, an individual can see what other users are
downloading from the individual's collection. This way, they are able to identify people with similar musical tastes and in turn peruse those people's collections.
This is sharing music with your friends, only on a much larger and widespread scale. Users are exposed to other musical possibilities, making this a step towards the center of the push-pull continuum.
The New Audio Consumer Consumer expectations
It was natural for initial content distribution over the Internet to essentially copy existing delivery paradigms, since they were most familiar to end users. Downloading a file was like owning it; you could store it and view/play it at your convenience. Conversely, streaming was akin to broadcasting, where content was not only delivered in real-time, but could only be experienced in real-time as well.
It became clear early on in the development of digital distribution channels that digital delivery technology offered much more potential than traditional delivery channels. The digital medium was more flexible and customizable and offered new possibilities. As new technologies were developed to take advantage of the opportunity, people were eager to embrace them and began to expect a certain amount of innovation and an increase in features and level of service. Of course, the most common access device to digital media is the personal computer, which is a device that puts a tremendous amount of control and power in the user's hands. Whether still at the computer or away from it, the consumers' expectation for the amount of control and power they ought to have over a delivery channel is now increased.
Consumers moving faster than industry
The migration from purely push models like broadcasting and pull models like the record industry towards some common point will see each of them tracing a path defined by technological opportunity and consumer demand. The pull industries like record labels are generally younger than broadcasters, and historically have been willing to redefine their formats (singles on 78s and 45s; collections on 33s; new editions on CDs) as newer technologies made these moves possible. And since the CD represented a move into the digital domain, the techniques of digital compression and efficient digital downloads were only a short step away-yet the record industry missed them both.
The essence of compressed file transfer is that both duplication and distribution of music suddenly becomes extremely cheap. Had the record industry embraced this and worked it into their business models, they could have instilled yet another format (in parallel with the more expensive CD), and had a chance to control both. If they had marketed singles on the Web for less than 50 cents a copy, the generally affluent PC owners would have most likely paid for them. This might have pulled CD prices down, but high-volume Web sales (at almost no cost) might have been able to sustain their profit margin. But instead,
the labels chose to stay with one technology, and to milk all they could from high CD prices. Enter mp3, and later Napster. Although the technology of mp3 has been around since the mid '90s, the audio encoding step was still significant at the time-except to a record producer whose Web market might reach millions. But once increased computer power had reduced the mp3 encode and decode cost to something trivial, consumers defined their own market model. And now Napster-organized file-sharing is suddenly the pull model to beat.
Push technology broadcasters have not yet been outmaneuvered in this way, but the potential is there. One enabling technology is caching , or the ability to store broadcast content for a short time and play it back when the consumer is ready. There is not yet a consumer compelling model that has significant market strength, although there are a few interesting examples. One of these is ClickRadio, which broadcasts to the Internet a predetermined playlist, interspersed with supporting commercials. The consumer does not listen to the broadcast in real-time, but to a delayed version cached on his computer. During listening, the consumer can elect to skip a song and move immediately to the next, or to the one following (there is no going backwards). After some number of songs, the consumer must hear a short commercial. This has economic viability by being only a slight variant on traditional, strictly push broadcast. However, variations that might have higher consumer appeal are not hard to dream up, e.g. play the cached songs in any requested order, or allow a cached song to be repeated up to 5 times (say) before it must be re-broadcast, etc. Moreover, if the player is clever about what you appear to favor, it could maximize the demographic advantage by playing only those commercials you are most likely to respond to. Similar options will accrue to streaming broadcast formats like digital radio, where several commercials can be multiplexed (broadcast simultaneously), so that the listener hears only the most relevant. This would maximize the demographic advantage by replacing simple blanket advertising with some- thing specific and more efficiently targeted. On the other hand, the recent growth of individual broadcasts, wherein users using SHOUTcast, Icecast and Myplay.com can put their custom mp3 playlists directly onto the net for others to hear, represents a grass-roots movement to redefine the entire Internet audio
entertainment space. These are essentially custom audio reflectors, and in one stroke they trivialize any distinction between push and pull.
Viable business models in these fast-changing areas will depend on two things, a timely sense of the opportunity
offered by technology, and an intellectual property system that is enlightened and moves with the times. Without these two working swiftly and in conjunction, the prospect for industries at both ends of the push-pull spectrum is that if the consumer moves to another technology first, the game is over. Then the traditional broadcasters and record labels can only play catch-up.
The Technologies Behind the Convergence
Advances in technology offer new and compelling possibilities to the consumer, which are accelerating the convergence of push and pull services. We believe
there are five technologies which will have the greatest impact on the future of audio entertainment.
-Audio coding
-Time-caching
-Network technology and infrastructure
-Collaborative filtering
-Interfaces
In the following sections, we explore
each of these technologies in more depth
to see how they will affect audio content
distribution services.
Audio coding
The technology of audio coding has been the driving force behind the digital music revolution, due to the widespread
acceptance of the mp3 standard (MPEG- 1, Layer 3). MPEG, the Motion Pictures Experts Group, is an international body
tasked with standardizing audio and video coding for storage and transmission. Mp3 is a compression technology
which stores digital music in approximately one-tenth the space of uncompressed music with little or no loss in
sound quality. It was standardized in 1992, but was only recently popularized with the advent of computers fast enough to handle the complex calculations of encoding and decoding music in real-time.
The technology of mp3 is known as perceptual sub-band coding, which uses knowledge about human hearing to essentially remove parts of the music we know to be inaudible. For example, a loud sound at one frequency will tend to obscure, or mask, quieter sounds of nearby frequencies. This phenomenon, called frequency masking, is the basis for mp3 and similar audio coding technologies. In this example, the quieter frequencies can be removed with little perceived difference in the sound, even though the actual sound data becomes quite different. The perceived difference is still there, however, and mp3 audio quality is noticeably inferior to an original CD recording. Other audio coders use the same basic technology and offer evolutionary improvements over mp3. The most recent coders such as MPEG-AAC (Advanced Audio Coding), Lucent e-PAC (enhanced Perceptual Audio Coding), Dolby AC-3, and Windows Media Audio offer another 50-100% of compression over mp3. Each of these is more complex than mp3 and requires a faster processor. However, sub-band technology has now reached its practical limit, and better compression will come only from a totally new direction.
Audio coding has historically gained in compression at the cost of computational complexity (Figure 2). A significant leap in this trend (both in compression and complexity) is a technology called Structured Audio, which is also part of the new MPEG-4 audio standard. Structured Audio, authored by researchers at the MIT Media Lab, incorporates novel technologies for synthetic audio. While synthetic sound has its limitations and certainly does not work for all situations, it allows for a huge amount of compression (100:1). This is because synthesis algorithms can be described very compactly, and computers are able to re-generate the audio from this symbolic description at the client end.
Structured Audio also allows synthetic audio to be synchronized with "naturally" (mp3-like) coded audio. An example of this would be to synchronize synthesized background instruments with a vocal track that is compressed using a "natural audio" coder. Overall, since most of the sound is synthetic, there is a dramatic savings in file size. Use of specialized compression algorithms (such as voice coders for vocal tracks) lead to further compression. Structured Audio achieves greater file compression (up to ten times greater) than mp3 by keeping
the various "tracks" of music (e.g. vocals, guitar, drums, synthesizer, etc.) separated prior to playback. During play-back, these tracks are reconstructed and then re-mixed to create the overall song.
This separated production and reassembly process, while complex, offers exciting possibilities for future music and
broadcast production. Since the song is re-mixed by the playback device, the device could choose an optimal mix for
the listening environment. For example, a living room with surround sound is best suited for a 5.1 Dolby Digital or DTS mix. A portable device with headphones would produce a headphone mix, and a car stereo would produce a 4-channel car-optimized mix. This technology isn't limited to music. Advertisement slots could be filled with several ads simultaneously, one of which could be selected by the playback device (see Interfaces below). In this way, Structured Audio technology (or something similar) has the potential to reshape the way music is experienced.
Time-caching
Computers and hard drives enable easy time-caching (short-term storage) of data. A well-established example of this
technology is seen in the the Personal Video Recorders (PVRs) sold by TiVo and ReplayTV. The most compelling use
of these devices is for time-shifting of content. These shifts can be as short as a few seconds or as long as needed. Short time-shifts are useful so that a user can
temporarily pause a "live" broadcast or begin a program a few minutes later to watch it in its entirety, e.g. arriving home five minutes after a program has begun.
Similar applications abound for audio: music broadcasts could be cached so they always begin at the start of a song; the broadcast could be paused to allow for interruptions; and appealing songs could be listened to again from the cache. Cached or stored content also
allows you to bypass less appealing parts of the broadcast, such as commercials. Just like with VCRs, most people with
PVRs skip the commercials. But while the PVR allows you to start watching a program just a few minutes after it's
begun, a VCR forces you to wait for its conclusion before viewing.
Of course, these new applications bring up certain legal and financial issues. Radio stations, based on sound
programming practices, don't repeat a song within short time intervals, so therefore, giving the consumer the ability to re-listen to a cached song (essentially copying it without permission) will certainly give pause to copyright holders. And if consumers can skip commercials, what is the point of advertising? These questions will require creative solutions. Copyright law will most likely need to evolve in order to be applicable to the digital age, and advertising will need to become more compelling to consumers.
A crucial point is that caching can easily become long-term storage. If most broadcasts are actually cached, and if
caches can become part of a user's permanent collection, the distinction between push (broadcast) and pull (adding to collection) becomes blurred. Caching technology is straightforward to implement, and even if devices don't yet
support the applications outlined here, they soon will. Early adopter consumers already expect and demand this amount of control and general consumers are well on their way to having those expectations. It is our view that, as with Napster, it will be impossible to reign in the technology, and it is in the industry's best interests to embrace and extend these technologies to offer compelling new services to consumers.
Network technology and infrastructure
Changes in the infrastructure of the Internet will accelerate the domination of digital music delivery. Mp3 files are a reasonable download at modem bandwidth, but still take several minutes to download. Widespread broadband access will allow downloads to occur much faster, the effects of which can be seen in the preponderance of mp3 files on the computers of college students (who generally have access to high-speed network
connections).
Perhaps more important than the spread of broadband access, however, are the advances in network quality of service. Web caching services are already increasing the quality of service of bandwidth-intensive content delivery. Services such as Akamai and the Real Broadcast Network store (cache) content on servers located throughout the world and attempt to deliver the requested content from the server closest to the end user. But even with broadband Internet access and web caching services, real-time streaming media is often unreliable and highly susceptible to interruptions in times of heavy network congestion. This is because all Internet data packets are treated equally, and time-critical data, such as streaming media, is indistinguishable from less time-critical data, such as e-mail. One solution to this problem is packet prioritization, where time-dependent packets are tagged with a higher priority than other packets. Higher priority packets will be forwarded first, decreasing the likelihood of interruption.
Another technique that will improve quality of service for streaming media is called multicasting. Currently, the
Internet consists of point-to-point communication, where one computer transmits data that is intended for only one
other computer. Streaming real-time media (such as a music broadcast) to a broader audience requires sending redundant data packets for each different listener (and computer). In multicasting, a single data packet can be forwarded to
many different receiving computers, reducing the amount of overall network traffic. Both packet prioritization and
multicasting require an upgrade of the communication protocols that are the foundation of the Internet. They are part of proposals included in the Internet2 initiative, which will be implemented at research universities over the next several years.
Another important area of growth in network technology is peer-to-peer (P-to-P) networking. Recently popularized by
such applications as Napster and Gnutella, the societal, behavioral, and financial impact of P-to-P is perhaps as
great (and as greatly feared) as any technology. While it efficiently enables sharing, distributing, and searching of content, less well-known is its potential as a technology to improve overall network efficiency. P-to-P combined with multicasting and distributed caching allows for much more efficient and resilient distribution of streaming content. These advances will allow for a much better streaming media experience and will further blur the distinction between push and pull services.
Internet traffic consists of many different types of data (communication, company data, digital media, etc.) Of course, one good possibility for dedicated digital
distribution of solely audio content is digital radio. Digital radio is already making inroads in Europe, where many stations simulcast their broadcasts in both analog and digital forms. Analog broadcast will eventually be phased out in several years once digital receivers become nearly ubiquitous. Of course digital radio offers many of the possibilities of Internet broadcast, but with a guarantee of service. The digital radio initiative in the US consists of terrestrial broadcasters and satellite service providers.
While the implementation of these technologies is inevitable, it must be noted that the pace of advances in network infrastructure have always lagged behind that of storage capacity and processing power. Thus, technologies have taken advantage of distributed storage and processing power first. The adoption of time-caching and new audio coding technologies will have a greater impact in a shorter time frame than eventual upgrades in bandwidth and quality of service.
Collaborative filtering
Digital content distribution offers possibilities for two-way communication that will change the way we obtain con-
tent and are exposed to advertising. Oftentimes, the search for new music content can be frustrating. New music
introduced over the radio is often criticized for being boring. And what about the thousands of recordings already in existence that a consumer hasn't yet listened to? There is a lot of great music out there to be found, and one very
important technology that addresses this need is
collaborative filtering.
Collaborative filtering creates matches between different individuals' preferences. For example, if Person A has sev-
eral favorite songs in common with Person B, they are likely to have similar musical preferences. So listening suggestions for Person A can be made from Person B's music collection and vice versa, as in the way friends of similar
musical tastes often make suggestions to each other. Collaborative filtering, however, can be much more powerful since your network of "friends" can be expanded to include the millions of music listeners on the Internet and better
matches and suggestions are much more likely to be made.
While advertising may still be needed to support some content distribution, it will need to become much more useful for the consumer or else it too will fall victim to technology (see Time-caching and Interfaces). The flip side of collaborative filtering allows for content distrib-
utors to highly target their customers. If a consumer's music preferences are known, marketing promotions can be
finely targeted to that individual's tastes. And if a user allows general demographic information to be made available, individually tailored (and potentially more
applicable) advertising becomes possible.
Interfaces
Our usual interfaces to audio entertainment are CD players, radios, and cassette players. However, computers and spe-cialized mp3 players (essentially small portable computers) are becoming increasingly common as music interfaces. Computers are the natural evolution for music devices since they easily store, manipulate, and replay large quantities of digital data, and they have the necessary logic to manage large collections of data. In general, computers have the benefit of being highly configurable and easily controlled, and most people have become accustomed to the power they wield when using a computer. As the primary interface to digital audio content, computers will encourage the consumer to expect and demand the same level of power in controlling their music environment as they do with their computing environment.
Computing devices also offer many potential options for the music content provider. Most mp3 players offer some
type of screen to display information, such as the song title, artist, album, etc. Digital radios also typically offer a comparable display. While useful as an interface to the device, the display could also be used to convey other information, particularly marketing and advertising. The devices could also track user information, preferences, and behavior so that advertising could be highly targeted. For example, if an advertisement slot on a digital radio broadcast actually included several possible ads, the digital radio could select which ad is most applicable to the user. In this way, no data is actually sent back to the broadcaster, so privacy issues aren't as much of a concern.
Trends for the Future
In summary, we believe that a combination of advancing technology and consumer demands and expectations will
drive audio entertainment services towards a common point combining aspects of both push and pull delivery modes.
Consumer expectations and demands:
-Consumers have needs for both push and pull content delivery services.
-Consumers expect a high degree of control over their listening experience.
-Consumers are quick to embrace new technologies and incorporate those into their expectations.
Advances in technology:
-Advances in audio coding, such as Structured Audio, will allow for more efficient and novel listening experiences
and offer new opportunities for content providers and distributors.
-Widespread use of time-caching removes some of the distinction between real-time and stored media, compensat-
ing for the lack of bandwidth and satisfying consumer demand for more control.
-Advances in networking technology will eventually allow efficient streaming and multicast, but not before other new
compensating technologies (coding and caching) take hold.
-Collaborative filtering provides new ways of finding new music and a new means of highly targeted advertising and
marketing.
-New interfaces will give the user large amounts of control, but will leave room for content providers to add or enhance services.
The convergence of push and pull services
What will this future world of converged push and pull services look like? On-demand streaming is sometimes put
forward as the solution for the future of audio entertainment. The argument is that with ubiquitous high-speed network access, local storage of content becomes unnecessary since consumers can simply stream any desired content. But we believe that this will not satisfy the desires of consumers who have come to expect more control over digital content. A simple illustration of this type of behavior is e-mail: people commonly connect to a server to check their e-mail, but download it to their computer instead of leaving it on the server. That way, the stored mail can be accessed even without a network connection. It can be processed (read, responded to, filed, and deleted) entirely off-line and then updated at the next connection. In a world of streaming-only delivery, losing your network connection would mean losing your ability to listen to music.
Instead we envision a device with a great deal of computational power and storage that can be connected and disconnected from a network (it may be wired or wireless). An early such device may simply be an mp3-recorder/player with a radio or wireless Internet tuner(s) that caches recently broadcast audio. The user should be able to go back and listen to recently played songs multiple times. The broadcast content may be advertiser supported, but advertisements will be multiplexed and the device will select ads appropriate for the user. Real-time broadcasts, such as sporting events, will be transmitted efficiently and without interruption. The device will be able to
connect to a search engine to find specific music, but will also have peer-to-peer capabilities to share music with other users and make intelligent suggestions from other users' playlists. The music itself may include bonus material (e.g. singles' flip-sides, or other song sugges-
tions). Permanent storage (purchase) of the music may require a small fee justified by the enhanced features of the music, such as optimized mixes for different sound environments, that offer an advantage over old mp3 files.
Other (even more far-fetched) possibilities for the future
Evolution of legislation to deal with copyright issues
The advances in technology and applications inevitably will lead to even greater disputes over intellectual property
and copyright. The technology has outgrown the current system, and content providers and distributors will need to
work to update copyright law.
Transformation of the audio entertainment industry
The convergence of push and pull services may lead to the convergence of the audio content and distribution
industries. With record labels looking to the Internet
for new channels to promote their music and with broadcasters looking to the Internet to expand their offerings to entice more listeners, it is conceivable that these once distinct industries may themselves find themselves consolidating into one.
Advice for the industry
-Seize the opportunity and find the ideal (for consumers) services point between push and pull before your com-
petitors do.
-Offer services that provide something more compelling to consumers than existing free (Napster, mp3) technologies.
-Look for ways of working with content creators and distributors to update copyright law.
http://www.interep.com/Radio2020/future.pdf
MIT Media Lab
September 5, 2000
The Nature of Push and Pull Delivery
The distribution of new media to customers has traditionally exploited two modes of delivery. The first mode is the broadcasting of content (such as radio or
television programs) to many users simultaneously. In this scenario, the end-user has little or no direct control over what content is being broadcast at any particular time (the only feedback being via ratings measurements of the audience for a particular program or station). The second
delivery mode is where the customer directly procures the content he or she desires, as with the purchase of a recording. Both modes of delivery have since embraced the world of digital delivery, where they have been coined push and
pull, respectively.
These two means of delivering content exist because we as consumers have a need for both. At times an individual will
know exactly what he or she wants to listen to, and pull is the best means of satisfying this consumer's needs. If we enjoy a particular song to the extent that we wish
to listen to it multiple times on multiple occasions, it makes sense to purchase that recording and simply play it whenever we wish. But there are also times when we
have no strong opinion about what we'd like to hear or only a general preference for a genre of music: perhaps some pop
music, maybe some jazz, or just something new. The broadcast (push) medium exists to fulfill this need. People will always have both these basic desires, and there will be occasion for content delivery in both modes, i.e.
there will always be a need for both push and pull.
The Convergence of Services Based on Push and Pull
Technological advances in digital media delivery are gradually moving push and pull services towards a common
point. As services built on the push mode become more mature, they start to incorporate features of the pull mode, and vice versa. We are witnessing the development
of a continuum that spans their difference, and the hitherto distinct services are beginning to populate points all along this spectrum. We expect future technology will enable a gravitating of services towards the mid-point of this spectrum, where features of both push and pull will
be equally important (Figure 1).
Figure 1: The continuum of push-pull services
Convergence
Unlimited sharing (Napster, Gnutella)
Basic user-control (ClickRadio TM )
"Micro-genres" (Internet radio)
Limited sharing (mp3)
CD purchase
Broadcast
Push oriented services
Push technology over the Internet began with the introduction of audio streaming, popularized by RealAudio.
While initially an attempt at content delivery on demand, the bandwidth limitations of the early Internet resulted in
low sound quality. However, its usefulness was quickly realized in applications such as speech (talk and sports radio), and thus streaming content delivery naturally adopted the traditional broadcast model. Radio stations would broadcast news and sporting events, and these broadcasts were not limited geographically, as they are with a traditional broadcast radio station. This was the primary force behind the early adoption of streaming: access to real-time out-of- market content.
As bandwidth and sound quality improved, music broadcasts became more prevalent on the Internet. Music stations began simulcasting their signal to the Internet, primarily so that out-of- market consumers could still listen to
their favorite stations. The most successful of these were highly specialized local stations, satisfying musical tastes not catered to outside of the local broadcast area. Now, this "micro-genrefication" has taken over Internet music broadcasting, with highly specialized stations, e.g. only 1983 British Pop, or all Charlie Parker. And while radio station simulcasts contain advertisements, many of the smaller Internet-only audio broadcasts do not. This provides an additional level of user control, not available over the traditional broadcast (push) medium, but moves the application towards the center of the push-pull continuum.
The next generation of push services introduce an even greater level of interactivity and control to the end user.
Services such as ClickRadioTM allow the user to skip a song they don't want to hear. These services are currently free
and are supported with advertising. And with greater interactivity comes more information about the user, which leads to better user profiling and the ability to finely target particular consumers.
Pull oriented services
The spread of pull technology for digital music delivery was catalyzed by the development of the mp3 standard for
audio compression. As soon as computers became powerful enough and storage space became cheap enough to make
playback and storage of mp3 files realistic, mp3 files became the standard method for aggregating digital music
content. Initially users would convert, or "rip", their own CDs into mp3 files on their computers. Even before sharing of files became widespread, many people converted their music libraries simply for the added convenience. This use of digital content represents the extreme end of user control. The music selection is entirely determined by the consumer's decisions regarding which music to rip.
Of course, the use of mp3 files exploded with the introduction of easy-to-use file-sharing software, most notably Napster. Prior to Napster, it was difficult to find and share popular music on the Internet. Large private collections of music existed, but they would appear and disappear in order to avoid being discovered and prosecuted by copyright enforcement agencies. With Napster, and other peer-to-peer networking systems, each user who logs into the system has the ability to share their entire collection. And as soon as they log out of the system, their collection disappears. Since each user only appears for a limited amount of time, it is difficult to ascertain which individuals are guilty of copyright infringement. However, since many (some fraction of over 20 million total) users are logged in at any particular time,
virtually any piece of music you desire can be found. It is a music lover's dream and a copyright holder's nightmare.
Napster, and its peer-to-peer brethren (Gnutella, Scour, and Freenet) have their bases in pull driven technologies, since the end user must specify which song he or she wishes to search for. However, these technologies represent a slightly different point in the push-pull continuum. As one searches on Napster, he or she is exposed to "serendipitous hits", or unintentional results of their search that may appeal to them. It may be a different recording or mix of the song they were looking for or an unfamiliar work by the same artist. Nevertheless, the user is exposed to additional music choices, that are presented to them. Additionally, an individual can see what other users are
downloading from the individual's collection. This way, they are able to identify people with similar musical tastes and in turn peruse those people's collections.
This is sharing music with your friends, only on a much larger and widespread scale. Users are exposed to other musical possibilities, making this a step towards the center of the push-pull continuum.
The New Audio Consumer Consumer expectations
It was natural for initial content distribution over the Internet to essentially copy existing delivery paradigms, since they were most familiar to end users. Downloading a file was like owning it; you could store it and view/play it at your convenience. Conversely, streaming was akin to broadcasting, where content was not only delivered in real-time, but could only be experienced in real-time as well.
It became clear early on in the development of digital distribution channels that digital delivery technology offered much more potential than traditional delivery channels. The digital medium was more flexible and customizable and offered new possibilities. As new technologies were developed to take advantage of the opportunity, people were eager to embrace them and began to expect a certain amount of innovation and an increase in features and level of service. Of course, the most common access device to digital media is the personal computer, which is a device that puts a tremendous amount of control and power in the user's hands. Whether still at the computer or away from it, the consumers' expectation for the amount of control and power they ought to have over a delivery channel is now increased.
Consumers moving faster than industry
The migration from purely push models like broadcasting and pull models like the record industry towards some common point will see each of them tracing a path defined by technological opportunity and consumer demand. The pull industries like record labels are generally younger than broadcasters, and historically have been willing to redefine their formats (singles on 78s and 45s; collections on 33s; new editions on CDs) as newer technologies made these moves possible. And since the CD represented a move into the digital domain, the techniques of digital compression and efficient digital downloads were only a short step away-yet the record industry missed them both.
The essence of compressed file transfer is that both duplication and distribution of music suddenly becomes extremely cheap. Had the record industry embraced this and worked it into their business models, they could have instilled yet another format (in parallel with the more expensive CD), and had a chance to control both. If they had marketed singles on the Web for less than 50 cents a copy, the generally affluent PC owners would have most likely paid for them. This might have pulled CD prices down, but high-volume Web sales (at almost no cost) might have been able to sustain their profit margin. But instead,
the labels chose to stay with one technology, and to milk all they could from high CD prices. Enter mp3, and later Napster. Although the technology of mp3 has been around since the mid '90s, the audio encoding step was still significant at the time-except to a record producer whose Web market might reach millions. But once increased computer power had reduced the mp3 encode and decode cost to something trivial, consumers defined their own market model. And now Napster-organized file-sharing is suddenly the pull model to beat.
Push technology broadcasters have not yet been outmaneuvered in this way, but the potential is there. One enabling technology is caching , or the ability to store broadcast content for a short time and play it back when the consumer is ready. There is not yet a consumer compelling model that has significant market strength, although there are a few interesting examples. One of these is ClickRadio, which broadcasts to the Internet a predetermined playlist, interspersed with supporting commercials. The consumer does not listen to the broadcast in real-time, but to a delayed version cached on his computer. During listening, the consumer can elect to skip a song and move immediately to the next, or to the one following (there is no going backwards). After some number of songs, the consumer must hear a short commercial. This has economic viability by being only a slight variant on traditional, strictly push broadcast. However, variations that might have higher consumer appeal are not hard to dream up, e.g. play the cached songs in any requested order, or allow a cached song to be repeated up to 5 times (say) before it must be re-broadcast, etc. Moreover, if the player is clever about what you appear to favor, it could maximize the demographic advantage by playing only those commercials you are most likely to respond to. Similar options will accrue to streaming broadcast formats like digital radio, where several commercials can be multiplexed (broadcast simultaneously), so that the listener hears only the most relevant. This would maximize the demographic advantage by replacing simple blanket advertising with some- thing specific and more efficiently targeted. On the other hand, the recent growth of individual broadcasts, wherein users using SHOUTcast, Icecast and Myplay.com can put their custom mp3 playlists directly onto the net for others to hear, represents a grass-roots movement to redefine the entire Internet audio
entertainment space. These are essentially custom audio reflectors, and in one stroke they trivialize any distinction between push and pull.
Viable business models in these fast-changing areas will depend on two things, a timely sense of the opportunity
offered by technology, and an intellectual property system that is enlightened and moves with the times. Without these two working swiftly and in conjunction, the prospect for industries at both ends of the push-pull spectrum is that if the consumer moves to another technology first, the game is over. Then the traditional broadcasters and record labels can only play catch-up.
The Technologies Behind the Convergence
Advances in technology offer new and compelling possibilities to the consumer, which are accelerating the convergence of push and pull services. We believe
there are five technologies which will have the greatest impact on the future of audio entertainment.
-Audio coding
-Time-caching
-Network technology and infrastructure
-Collaborative filtering
-Interfaces
In the following sections, we explore
each of these technologies in more depth
to see how they will affect audio content
distribution services.
Audio coding
The technology of audio coding has been the driving force behind the digital music revolution, due to the widespread
acceptance of the mp3 standard (MPEG- 1, Layer 3). MPEG, the Motion Pictures Experts Group, is an international body
tasked with standardizing audio and video coding for storage and transmission. Mp3 is a compression technology
which stores digital music in approximately one-tenth the space of uncompressed music with little or no loss in
sound quality. It was standardized in 1992, but was only recently popularized with the advent of computers fast enough to handle the complex calculations of encoding and decoding music in real-time.
The technology of mp3 is known as perceptual sub-band coding, which uses knowledge about human hearing to essentially remove parts of the music we know to be inaudible. For example, a loud sound at one frequency will tend to obscure, or mask, quieter sounds of nearby frequencies. This phenomenon, called frequency masking, is the basis for mp3 and similar audio coding technologies. In this example, the quieter frequencies can be removed with little perceived difference in the sound, even though the actual sound data becomes quite different. The perceived difference is still there, however, and mp3 audio quality is noticeably inferior to an original CD recording. Other audio coders use the same basic technology and offer evolutionary improvements over mp3. The most recent coders such as MPEG-AAC (Advanced Audio Coding), Lucent e-PAC (enhanced Perceptual Audio Coding), Dolby AC-3, and Windows Media Audio offer another 50-100% of compression over mp3. Each of these is more complex than mp3 and requires a faster processor. However, sub-band technology has now reached its practical limit, and better compression will come only from a totally new direction.
Audio coding has historically gained in compression at the cost of computational complexity (Figure 2). A significant leap in this trend (both in compression and complexity) is a technology called Structured Audio, which is also part of the new MPEG-4 audio standard. Structured Audio, authored by researchers at the MIT Media Lab, incorporates novel technologies for synthetic audio. While synthetic sound has its limitations and certainly does not work for all situations, it allows for a huge amount of compression (100:1). This is because synthesis algorithms can be described very compactly, and computers are able to re-generate the audio from this symbolic description at the client end.
Structured Audio also allows synthetic audio to be synchronized with "naturally" (mp3-like) coded audio. An example of this would be to synchronize synthesized background instruments with a vocal track that is compressed using a "natural audio" coder. Overall, since most of the sound is synthetic, there is a dramatic savings in file size. Use of specialized compression algorithms (such as voice coders for vocal tracks) lead to further compression. Structured Audio achieves greater file compression (up to ten times greater) than mp3 by keeping
the various "tracks" of music (e.g. vocals, guitar, drums, synthesizer, etc.) separated prior to playback. During play-back, these tracks are reconstructed and then re-mixed to create the overall song.
This separated production and reassembly process, while complex, offers exciting possibilities for future music and
broadcast production. Since the song is re-mixed by the playback device, the device could choose an optimal mix for
the listening environment. For example, a living room with surround sound is best suited for a 5.1 Dolby Digital or DTS mix. A portable device with headphones would produce a headphone mix, and a car stereo would produce a 4-channel car-optimized mix. This technology isn't limited to music. Advertisement slots could be filled with several ads simultaneously, one of which could be selected by the playback device (see Interfaces below). In this way, Structured Audio technology (or something similar) has the potential to reshape the way music is experienced.
Time-caching
Computers and hard drives enable easy time-caching (short-term storage) of data. A well-established example of this
technology is seen in the the Personal Video Recorders (PVRs) sold by TiVo and ReplayTV. The most compelling use
of these devices is for time-shifting of content. These shifts can be as short as a few seconds or as long as needed. Short time-shifts are useful so that a user can
temporarily pause a "live" broadcast or begin a program a few minutes later to watch it in its entirety, e.g. arriving home five minutes after a program has begun.
Similar applications abound for audio: music broadcasts could be cached so they always begin at the start of a song; the broadcast could be paused to allow for interruptions; and appealing songs could be listened to again from the cache. Cached or stored content also
allows you to bypass less appealing parts of the broadcast, such as commercials. Just like with VCRs, most people with
PVRs skip the commercials. But while the PVR allows you to start watching a program just a few minutes after it's
begun, a VCR forces you to wait for its conclusion before viewing.
Of course, these new applications bring up certain legal and financial issues. Radio stations, based on sound
programming practices, don't repeat a song within short time intervals, so therefore, giving the consumer the ability to re-listen to a cached song (essentially copying it without permission) will certainly give pause to copyright holders. And if consumers can skip commercials, what is the point of advertising? These questions will require creative solutions. Copyright law will most likely need to evolve in order to be applicable to the digital age, and advertising will need to become more compelling to consumers.
A crucial point is that caching can easily become long-term storage. If most broadcasts are actually cached, and if
caches can become part of a user's permanent collection, the distinction between push (broadcast) and pull (adding to collection) becomes blurred. Caching technology is straightforward to implement, and even if devices don't yet
support the applications outlined here, they soon will. Early adopter consumers already expect and demand this amount of control and general consumers are well on their way to having those expectations. It is our view that, as with Napster, it will be impossible to reign in the technology, and it is in the industry's best interests to embrace and extend these technologies to offer compelling new services to consumers.
Network technology and infrastructure
Changes in the infrastructure of the Internet will accelerate the domination of digital music delivery. Mp3 files are a reasonable download at modem bandwidth, but still take several minutes to download. Widespread broadband access will allow downloads to occur much faster, the effects of which can be seen in the preponderance of mp3 files on the computers of college students (who generally have access to high-speed network
connections).
Perhaps more important than the spread of broadband access, however, are the advances in network quality of service. Web caching services are already increasing the quality of service of bandwidth-intensive content delivery. Services such as Akamai and the Real Broadcast Network store (cache) content on servers located throughout the world and attempt to deliver the requested content from the server closest to the end user. But even with broadband Internet access and web caching services, real-time streaming media is often unreliable and highly susceptible to interruptions in times of heavy network congestion. This is because all Internet data packets are treated equally, and time-critical data, such as streaming media, is indistinguishable from less time-critical data, such as e-mail. One solution to this problem is packet prioritization, where time-dependent packets are tagged with a higher priority than other packets. Higher priority packets will be forwarded first, decreasing the likelihood of interruption.
Another technique that will improve quality of service for streaming media is called multicasting. Currently, the
Internet consists of point-to-point communication, where one computer transmits data that is intended for only one
other computer. Streaming real-time media (such as a music broadcast) to a broader audience requires sending redundant data packets for each different listener (and computer). In multicasting, a single data packet can be forwarded to
many different receiving computers, reducing the amount of overall network traffic. Both packet prioritization and
multicasting require an upgrade of the communication protocols that are the foundation of the Internet. They are part of proposals included in the Internet2 initiative, which will be implemented at research universities over the next several years.
Another important area of growth in network technology is peer-to-peer (P-to-P) networking. Recently popularized by
such applications as Napster and Gnutella, the societal, behavioral, and financial impact of P-to-P is perhaps as
great (and as greatly feared) as any technology. While it efficiently enables sharing, distributing, and searching of content, less well-known is its potential as a technology to improve overall network efficiency. P-to-P combined with multicasting and distributed caching allows for much more efficient and resilient distribution of streaming content. These advances will allow for a much better streaming media experience and will further blur the distinction between push and pull services.
Internet traffic consists of many different types of data (communication, company data, digital media, etc.) Of course, one good possibility for dedicated digital
distribution of solely audio content is digital radio. Digital radio is already making inroads in Europe, where many stations simulcast their broadcasts in both analog and digital forms. Analog broadcast will eventually be phased out in several years once digital receivers become nearly ubiquitous. Of course digital radio offers many of the possibilities of Internet broadcast, but with a guarantee of service. The digital radio initiative in the US consists of terrestrial broadcasters and satellite service providers.
While the implementation of these technologies is inevitable, it must be noted that the pace of advances in network infrastructure have always lagged behind that of storage capacity and processing power. Thus, technologies have taken advantage of distributed storage and processing power first. The adoption of time-caching and new audio coding technologies will have a greater impact in a shorter time frame than eventual upgrades in bandwidth and quality of service.
Collaborative filtering
Digital content distribution offers possibilities for two-way communication that will change the way we obtain con-
tent and are exposed to advertising. Oftentimes, the search for new music content can be frustrating. New music
introduced over the radio is often criticized for being boring. And what about the thousands of recordings already in existence that a consumer hasn't yet listened to? There is a lot of great music out there to be found, and one very
important technology that addresses this need is
collaborative filtering.
Collaborative filtering creates matches between different individuals' preferences. For example, if Person A has sev-
eral favorite songs in common with Person B, they are likely to have similar musical preferences. So listening suggestions for Person A can be made from Person B's music collection and vice versa, as in the way friends of similar
musical tastes often make suggestions to each other. Collaborative filtering, however, can be much more powerful since your network of "friends" can be expanded to include the millions of music listeners on the Internet and better
matches and suggestions are much more likely to be made.
While advertising may still be needed to support some content distribution, it will need to become much more useful for the consumer or else it too will fall victim to technology (see Time-caching and Interfaces). The flip side of collaborative filtering allows for content distrib-
utors to highly target their customers. If a consumer's music preferences are known, marketing promotions can be
finely targeted to that individual's tastes. And if a user allows general demographic information to be made available, individually tailored (and potentially more
applicable) advertising becomes possible.
Interfaces
Our usual interfaces to audio entertainment are CD players, radios, and cassette players. However, computers and spe-cialized mp3 players (essentially small portable computers) are becoming increasingly common as music interfaces. Computers are the natural evolution for music devices since they easily store, manipulate, and replay large quantities of digital data, and they have the necessary logic to manage large collections of data. In general, computers have the benefit of being highly configurable and easily controlled, and most people have become accustomed to the power they wield when using a computer. As the primary interface to digital audio content, computers will encourage the consumer to expect and demand the same level of power in controlling their music environment as they do with their computing environment.
Computing devices also offer many potential options for the music content provider. Most mp3 players offer some
type of screen to display information, such as the song title, artist, album, etc. Digital radios also typically offer a comparable display. While useful as an interface to the device, the display could also be used to convey other information, particularly marketing and advertising. The devices could also track user information, preferences, and behavior so that advertising could be highly targeted. For example, if an advertisement slot on a digital radio broadcast actually included several possible ads, the digital radio could select which ad is most applicable to the user. In this way, no data is actually sent back to the broadcaster, so privacy issues aren't as much of a concern.
Trends for the Future
In summary, we believe that a combination of advancing technology and consumer demands and expectations will
drive audio entertainment services towards a common point combining aspects of both push and pull delivery modes.
Consumer expectations and demands:
-Consumers have needs for both push and pull content delivery services.
-Consumers expect a high degree of control over their listening experience.
-Consumers are quick to embrace new technologies and incorporate those into their expectations.
Advances in technology:
-Advances in audio coding, such as Structured Audio, will allow for more efficient and novel listening experiences
and offer new opportunities for content providers and distributors.
-Widespread use of time-caching removes some of the distinction between real-time and stored media, compensat-
ing for the lack of bandwidth and satisfying consumer demand for more control.
-Advances in networking technology will eventually allow efficient streaming and multicast, but not before other new
compensating technologies (coding and caching) take hold.
-Collaborative filtering provides new ways of finding new music and a new means of highly targeted advertising and
marketing.
-New interfaces will give the user large amounts of control, but will leave room for content providers to add or enhance services.
The convergence of push and pull services
What will this future world of converged push and pull services look like? On-demand streaming is sometimes put
forward as the solution for the future of audio entertainment. The argument is that with ubiquitous high-speed network access, local storage of content becomes unnecessary since consumers can simply stream any desired content. But we believe that this will not satisfy the desires of consumers who have come to expect more control over digital content. A simple illustration of this type of behavior is e-mail: people commonly connect to a server to check their e-mail, but download it to their computer instead of leaving it on the server. That way, the stored mail can be accessed even without a network connection. It can be processed (read, responded to, filed, and deleted) entirely off-line and then updated at the next connection. In a world of streaming-only delivery, losing your network connection would mean losing your ability to listen to music.
Instead we envision a device with a great deal of computational power and storage that can be connected and disconnected from a network (it may be wired or wireless). An early such device may simply be an mp3-recorder/player with a radio or wireless Internet tuner(s) that caches recently broadcast audio. The user should be able to go back and listen to recently played songs multiple times. The broadcast content may be advertiser supported, but advertisements will be multiplexed and the device will select ads appropriate for the user. Real-time broadcasts, such as sporting events, will be transmitted efficiently and without interruption. The device will be able to
connect to a search engine to find specific music, but will also have peer-to-peer capabilities to share music with other users and make intelligent suggestions from other users' playlists. The music itself may include bonus material (e.g. singles' flip-sides, or other song sugges-
tions). Permanent storage (purchase) of the music may require a small fee justified by the enhanced features of the music, such as optimized mixes for different sound environments, that offer an advantage over old mp3 files.
Other (even more far-fetched) possibilities for the future
Evolution of legislation to deal with copyright issues
The advances in technology and applications inevitably will lead to even greater disputes over intellectual property
and copyright. The technology has outgrown the current system, and content providers and distributors will need to
work to update copyright law.
Transformation of the audio entertainment industry
The convergence of push and pull services may lead to the convergence of the audio content and distribution
industries. With record labels looking to the Internet
for new channels to promote their music and with broadcasters looking to the Internet to expand their offerings to entice more listeners, it is conceivable that these once distinct industries may themselves find themselves consolidating into one.
Advice for the industry
-Seize the opportunity and find the ideal (for consumers) services point between push and pull before your com-
petitors do.
-Offer services that provide something more compelling to consumers than existing free (Napster, mp3) technologies.
-Look for ways of working with content creators and distributors to update copyright law.
http://www.interep.com/Radio2020/future.pdf
Join the InvestorsHub Community
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
