Blank (Media) Slate

Those of you following the story of my new business venture, matchmine, know that there’s a part of it that came from discussions my brother and I had back in ‘95. Specifically, we were talking about the implications of music jumping off their prison of plastic discs and being more freely distributed online as bits and bytes. We knew the paradigm shift was coming, and we worked out some concepts for what that’d mean to end users.

Around that time Mark was working on an open source project with some MIT alum to build a digital jukebox. Keep in mind, though, that this was before “MP3″ was a household term (I even remember some debate raging as to which format it’d support). Using some nifty AI tricks, the jukebox was able to cleverly serve up a good mix of tracks that’d largely hit the interests of the folks connected to it. Amusingly, this whole project was primarily driven by the desire to help his staff listen to their collective tunes.

That was then, and this is now. And now I just fielded a call from him with another clever idea, this time including the artists in the mix. The basic concept is rooted in the fact that the music industry as it has been for the past fifty years will have to change to remain relevant. There’s a lot of debate about the actual form the new model will eventually take, but the dust isn’t close to settling on the right solution just yet. In fact, unlike the monolithic solutions of the past, it’s distinctly possible that multiple models can simultaneously exist.

To that end, the idea du jour is how to take the digital jukebox concept and pay the artists directly when people listen. As it stands now, for any one artist there are all sorts of contracts with multiple parties (eg. producers, labels, distributors, etc.) the majority of which are governed by the RIAA. It’s obvious why that has historically proven to be “the way of it”, but the rules of the game have changed significantly enough in recent years that it’s worth evaluating new opportunities.

With this recent conversation rattling in my head, I stumbled onto a post by Jason Kolb that sketches a basic concept speaking to the plan Mark and I discussed:

One of my friends was handing out CD’s for a band that his friend was a part of, which got me thinking. If people are willing to hand out CD’s for bands because they like them, why not capitalize on that and turn it into an associate program? It would be cheaper and easier to just hand out cards with some information including a URL to get free MP3 downloads from the Web site. The Web site could offer the rest of the tracks as a paid download along with selling merchandise and tickets—the cash cow. Just put a unique identifier in the URL so you can track it back to the person who gave the card out, and give them a cut of any revenue that comes in as a result of their evangelism. All the benefit of record label marketing without the gatekeepers and middlemen.

Formalizing the idea, then, and without further adieu here’s basically what Mark suggested — to build a tool that an end user can:

• Identify yourself (either directly or by representation)
• Listen to music you’d like (either by parameter command, or based on a preference profile)
• Get track, album, artist information (so you can learn more about them)
• Action on purchase events (such as buying tracks, albums, concert tickets, or “tipping” the artist directly)
• Access standard social networking tools (that are too obvious at this stage to enumerate)

This could be in the form of a web site… though I’d suggest a simple web service that other websites and applications could use. To do so, some light-weight API would do the trick. There are already too many end user destinations as it is; why add to the noise? This approach would also allow existing (or new) destinations to generate some revenue of their own with a piece of advertising, etc.

All very well-and-good, but how’s the music gonna’ get in there to be fed back out? The answer is that the music catalog only contains tracks released under a generally compatible open license (eg. Creative Commons). A lot of which can already be obtained online under various licenses, but these specific artists never expected to see a dime off of these tracks. To encourage wider support, though, they’d need something more than nothing for their work. To service them, another tool set would need to be rolled out for the artists:

• Manage your identity and public info (by legal name or by pseudonym)
• Manage your tracks (upload, modify, retire, etc.)
• Set the license by which the tracks are released (from a closed set of compatible licenses)
• Select an economic model (eg. free, pay-per-listen, pay-per-download, downstream revenue percentage, etc.)

These tools would probably need to be coupled with similar ones geared toward downstream revenue generators. It’d be reasonable to assume something like what Kolb mentions could be the simple first step (ie. referral service to direct purchase). Following that, additional services can be added to the portal and associated APIs over time as it builds momentum.

Anyway, there’s a brief sketch of the new thought. Given the accelerating pace, I’m guessing it won’t be long to see something like it come to fruition. In fact… at matchmine we’re already building [CENSORED BY MATCHMINE DEPARTMENT OF SECRECY] and we’re really excited about it!

I’ve been reading up on memory lately and found an interesting effect that might speak (in a totally non-scientific way) to something that seems to be part of our collective culture. I was listening to research CDs by Diana Deusch (see an earlier post for more background) and a pitch memory experiment caught my attention.

Here’s a quick blurb on “Short Term Memory for Pitch” from her site:

Suppose you play a tone, and this is followed by another tone which is either the same in pitch as the first, or differs by a semitone, most people find it very easy to decide whether the two tones are the same or different. This is true even when the tones are separated by a five second delay.

That’s all very well and good, of course, but when the intervening silence is filled with other tones, the task becomes much more difficult. Nothing strange there, I think we all get that. What gets more interesting, though, is when you fill the space between tones with spoken words (in this case they are spoken numbers). As Deusch says:

Most people find the memory task much easier when spoken numbers rather than tones are played during the interval between the test tones. This contrasts remarkably with the memory loss that occurs when a sequence of tones is played during the interval between the test tones, even though the extra tones can be ignored. So we conclude that the pitch of a tone is held in a specialized memory store, and that interference takes place between pitches inside this store. Other materials - such as spoken numbers - do not enter the store, so they produce much less interference with memory for pitch.

According to the background reading I’ve done in memory, it turns out that short term memory can only juggle a handful of similar items before they are either committed to longer term memory or lost. In fact, George Miller, while at Bell Labs, determined that the number of similar items is about seven, plus/minus two. Research has since shown that the number of “storable items” differs by type and situation (eg. seven for digits, six for letters, and five for words).

What I find interesting is how this might map into popular culture. For example, I’ve always wondered why people tend to remark about recent films being “the best movie ever.” I don’t take those statements literally, but assume them to be infused with exuberant hyperbole to make a point. It’s also relatively well known that in order to capitalize on this effect, movies in contention for Oscars are purposefully released near voting time.

So, if all this is well known, what’s the point of this post? First of all, I think it’s interesting to play with the temporal effect of “liking” something (ie. adding more relative weight to recent interests, degrading them over time). Second, I think it’d be interesting to explore if we’re hardwired to only allow for something like 7 +/- 2 “greatest” movies/songs/etc. in our heads at any given time.

I found an interesting new (beta launched 2/13/2007) search engine called hakia. From the looks of it, they’re rolling their own solution on semantic-based stuff. Here’s a blurb from their site:

The basic promise is to bring search results by meaning match - similar to the human brain’s cognitive skills - rather than by the mere occurrence (or popularity) of search terms. hakia’s new technology is a radical departure from the conventional indexing approach, because indexing has severe limitations to handle full-scale semantic search.

Interestingly, they purposefully call out specific uses in which they believe their solution is particularly well-suited:

hakia’s capabilities will appeal to all Web searchers - especially those engaged in research on knowledge intensive subjects, such as medicine, law, finance, science, and literature.

I hammered on it for a bit, and it does look like it’s got some good feet under it. I’ll try replacing it as my go-to search site for a while and see how it goes (similar to what I did with AltaVista when I found Google in 1997 - never to look back). More on the experiment - if it develops.

I turned up a short counter-point blog post about their approach by Marc Fawzi and ToxicWave:

“We are beginning to see search engines that claim they can semantic-ize arbitrary unstructured “Wild Wild Web” information. Wikipedia pages, constrained to the Wikipedia knowledge management format, may be easier to semantic-ize on the fly. However, at this early stage, a better approach may be to use human-directed crawling that associates the information sources with clearly defined domains/ontologies.”

I like that idea… at least until the machines are smart enough to push aside their masters (as anyone who reads science fiction knows they’ll do eventually).

Over the past year I’ve been focused on designing a system for effectively identifying media content based on personal interests. Throughout multiple discussions over that time, I’ve often stressed the need to embrace (rather than eradicate) the value of serendipity in results. Two comments in the past week have really driven that point home and given this idea a voice I find useful.

The first comes from our executive leadership in recasting the system in terms of a “Discovery Platform.” While it might seem to be a minor semantic variance, the term “discovery” fits nicely within our context. We’ve taken great pains to distance ourselves from the traditional and, IMHO, limiting tradition of finding and delivering the “perfect result” within the specified candidate pool. Rather, I’ve generally used the term “serendipity” to evoke the experiential feeling we hope to provide. I still like the feel of the term, though it’s heavy connotation of “chance” puts it second to a more directed term like “discovery”, and it still embraces the concept of finding hitherto “unknown” matches for your interests.

The second comment came in meetings with a potential product developer who said, “Effective recommendations must be relevant, but not predictable.” It’s possible this is a well-trodden phrase, but it resonated with our solidifying vision around discovery. After all, it’s easy to recommend songs by the same artist you already like (or following other paths similarly codified), it takes something more to uncover less obvious connections to your interests. Further, it’s a lot more gratifying (like the punch line of a good joke) when you’re pleasantly surprised by the result.

As a side note, a common refrain around the office has been, “You don’t discover pocket lint.” I’m still not entirely sure I grok the true meaning from this, but it definitely pops an amusing image in your mind’s eye.