Web sites where customers can legally download music are booming. Since its launch in October 2003, Apple’s iTunes Music Store has sold 200 million songs. Following in those footsteps, a group of scientists headed by experts at CSIC, the center for advanced scientific research at the Autonomous University of Barcelona, Spain, have gone one step further. They have created MusicStrands, a company that uses artificial intelligence to recommend music over the Net.
“Until iTunes appeared, there were only peer-to-peer sites, such as Kazaa, eMule and eDonkey, where users could download songs totally free. There were also sites that sold entire albums by artists,” explains Xabier Ruiz de Aguirre, vice-president of business development of MusicStrands, a new sort of music portal. Not only did iTunes challenge those illegal download sites, it also managed to achieve an overwhelming 70% share of the market. Microsoft, Yahoo, Wanadoo and Terra followed by launching their own online music stores.
The Google of Music
MusicStrands, dubbed “The Google of music,” began operations on February 9. However, “the plans were conceived three years ago, when iTunes appeared,” says Ruiz de Aguirre. “We saw that music was a business that had great potential.” MusicStrands does not sell music. Instead, it is a web site that helps users find songs and other news that fit their tastes, using artificial intelligence tools.
The concept behind MusicStrands came from Francisco J. Martín, the Spaniard who was previously founder and managing director of iSoco, an electronic commerce company that makes software called “intelligent agents.” To found MusicStrands, Martin brought together scientists from Switzerland and the United States to develop an innovative technology based on artificial intelligence. The new project was conceived in the Artificial Intelligence Institute of the CSIC, Spain’s center for advanced scientific research, at the Autonomous University of Barcelona. They also got help from scientists at the Artificial Intelligence Laboratory of the Federal Institute of Technology in Lausanne, Switzerland and from the Department of Electrical Engineering and Computer Science at Oregon State University.
They created a team consisting of 22 people. It included John Herlocker, adjunct professor of electrical engineering and computer science at Oregon State University; Derek Reisfield, ex-president of CBS New Media; Tom Dietterich, president of the International Machine Learning Society; Javier Etxebeste, ex-director of research for Yahoo’s European marketplace, and Andreas Weigend, former chief scientist at Amazon.com. This innovative company was launched with initial capital of $1 million (€750,000), which came from its management and private investors. MusicStrands is now trying to raise additional capital from institutional investors.
The Business Model
Here is how the MusicStrands portal works: The MusicStrands web site, at www.musicstrands.com, contains a catalog of 3.7 million songs in a wide range of song formats, disc formats, computer software programs, and listening devices. When an individual visits this portal, he or she uses a search engine to search that catalog either by the name of the artist, the title of the album or the title of the song. “We wanted to create a simple, fast search engine, since it is not easy to navigate three million titles, not to mention [trying to] find what you want right away,” says Ruiz de Aguirre. “In addition to exploring the catalog, users can listen to 30 seconds of each song at no cost, and create ‘play lists’ of their favorite songs, as well as ‘wish lists’ of other songs. They can also find new music, using a recommendation engine.”
That tool, known as the Recommender, is truly the soul of the company. Its mission is to analyze the tastes of each user, using emotional intelligence technology. “This system makes recommendations based on what the user reads, listens to, and sees,” notes Ruiz de Aguirre. “This technology is able to detect how many times a user listens to a specific song, and it compares that data with other songs that he or she listens to, in order to determine his or her preferences. Moreover, the tastes of the user are compared with the behavior of other users around the world who have a similar style. Based on those customers, MusicStrands can make very reliable recommendations to the user.” It works a lot like the Google search engine: Everything is automatic, and there is no need for human intervention.
John Herlocker, general manager of MusicStrands, developed the Recommender. “Herlocker already had filed patents for Amazon’s recommendation engine,” says Ruiz de Aguirre. “It recommends songs, books, and movies based on users’ preferences, and those of customers who make the same purchases. However, the Amazon recommendation agent has several defects. For example, it requires users to interact [with the site]; Amazon asks each buyer to explain how much he or she likes a new purchase. Also, Amazon does not know whether a purchase is for the user or for another person.”
The name of the company, MusicStrands, sums up its philosophy. The word “strands” can be interpreted metaphorically to mean musical DNA. Thus, the project is fundamentally involved in identifying musical tastes, or discovering each user’s unique musical DNA. However, this process takes place without forgetting that recommendations are always changing because they are based on changing tastes. “Strands” also means trends, or currents.
So far, the scientific team at MusicStrands has obtained five patents. It expects to obtain up to 40 patents over the next five years. By 2007, the company forecasts a total investment volume of more than 60 million euros ($79 million). That will enable MusicStrands to pay its suppliers, keep developing new technology, and maintain its online platform. By 2010, the company expects to achieve revenues of $60 million and earnings of $20 million.
Sources of Revenue
MusicStrands does not sell the music itself, and its end-users do not pay anything for the service that MusicStrands provides. Instead, the role of MusicStrands is to redirect its users to other web portals where they can buy their songs. With that goal in mind, MusicStrands has reached agreements with Amazon, Napster, Apple, and Disco Web. “We are an independent platform, so we can provide complete catalogs of all discs that are available,” says Ruiz de Aguirre. “We also want to promote independent labels, including their songs in our catalog. We want to redirect users to those companies’ portals. What we do not recommend, however, are web pages where people can illegally download music.”
So how does MusicStrands make money? “Each time we redirect a user to one of the pages of our partner companies, we earn between 5% and 10% of the final sale price.” There will also be another revenue stream in the future. “The value of the company comes from its users, because we can use the personal information that they give us,” says Ruiz de Aguirre. “We can sell that information to the disc companies and marketers. We can also create reports for the industry that show which kinds of music people are listening to nowadays. That way, people can make [informed] predictions about what is going to be a hit in coming years.”
In the future, he adds, “we want to start making money through advertising. We want to offer sponsored links, much like those used by Google, but always related to the world of music.”
The MusicStrands team, which has offices in Barcelona and [Corvallis,] Oregon, is looking into the possibility of synchronizing its online Recommender with MP3 players, Apple’s iPod, the Discman, and personal computers. That way, says Ruiz de Aguirre, “our web site would be able to recognize the songs that each user is listening to. It could identify them and learn more about his or her preferences. This new service will arrive in four months.” In addition, “in June, we will introduce a service for managing music libraries. This system consists of a tool, based on graphical music libraries, that allows people to create a simpler system. The new technology is based on scientific research by Marc Torrens, a Spaniard, who is vice-president of technology at MusicStrands. Torrens is a specialist in intelligence systems and electronic catalogs.”
Ultimately, MusicStrands does not want to limit itself to the realm of online music. “In the future, we will take our Recommender to mobile telephony and to digital cable television,” Ruiz de Aguirre says. In the future, MusicStrands’ recommendation agent will be purchased by many companies, and it will be a global standard.