[jahsonic.com] - [Next >>]

Jim Carrico


Sacramento, CA-based graphic artist

[potlatch] Re: pho: Let It Be 
Jim Carrico jim@at.org 
Mon, 20 Aug 2001 12:11:31 -0700 

Previous message: : Re: [potlatch] Watermarking 
Next message: [potlatch] Re:  Garcia/Lovink: The GHI of Tactical Media 
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] 


(re-post from the Pho list)

Michael Papish  wrote:
>there is no correlation between the amount of information you know about
>the listener and the depth of obscurity you can achieve in recommending
>song.  if fact, one of the most useful data points is not "a rating," but
>how open the listener is in hearing "unfamilar" music.
>i believe you are using the term "limited-data model" incorrectly. what
>you are talking about is having limited preference data on obscure
>artists/songs from the population as a whole (not limited "ratings" data
>from an individual). in this case, you are correct about the failure of
>straight Collaborative Filtering or any other computer-only algorithim.
>this is why Pattie Maes says the following about pure CF-based 
>Firefly: "One big problem that Firefly didn't solve is that when all 
>recommendations are only based on what other people enter, then it's
>harder for new artists to get recommended. Enough other people need to
>have taken an initiative to put in a rating for an artist before it gets
>recommended to people with similar tastes."
>you can solve this by blending the CF and other computer-based technology
>with a small group of human music analysts who help tweak the system by
>creating relational maps, adding and promoting obscure artists and
>devising ways of being more intelligent in gathering information from
>listeners who don't want to sit around and rate songs.

unfortunately, the "small group of human music analysts who help 
tweak the system" sounds a bit too much like the "independant 
promoters" who mysteriously  influence the existing 
popularity indices.  If we were to imagine a system of discovering 
and promoting unknown artists that does not require massive 
promotional budgets, and is not susceptible to manipulation and shady 
practises, it cannot rely on the good taste or good intentions of a 
"small group".  In general terms, a single point of failure in a 
complex system is bad design.

Right now, in order to be popular, music needs to have a certain 
innate "quality " to it - lets call it a resonance with a significant 
cross-section of the population - but it equally must have exposure 
to the ears of this same group.  For example, although folk/trad 
artists such as Alison Krauss and the Cox family have been plying 
their trade in relative obscurity for years, it took the Coen 
Brothers to bring them to a mainstream audience with "Oh brother, 
where art thou".  This must be some kind of smash hit,  seems like I 
see the soundtrack CD everywhere.  The fact is, it's great music, it 
sounds fresh, and it would *never* have found its audience without 
going "out of band" and routing around the RIAA hit machine. The 
point is, there's a lot of great music out there that could be 
popular but isn't, simply because most people have never been exposed 
to it - the existing hit-based promo system  *isn't working*.

So how to design a promotional system that doesn't depend on a 
centralized (and therefore vulnerable) taste-making agency?  A few 
months ago Ian Clarke of Freenet broadcast a suggestion for creating 
incentives for voluntary payments to artists and creators, which he 
called "Fairshare".  The basic idea is that payments should be shared 
equally between the creators and the previous "payers", essentially a 
decentralization/democratization of investment. Of course this sounds 
suspiciously like a pyramid scheme - but rather than concentrating 
profits "up-line", it tends to spread them thinner and thinner as the 
pyramid widens.  (It works out that a payment will be repaid by the 
time the pyramid is roughly 7 times bigger than it was when the 
payment was made.)

No one seems to have taken up this idea, but I find myself returning 
to it when thinking about alternative promotional mechanisms.  It 
seems to create incentives for individuals to "invest" early and then 
promote the hell out of whatever it is they've invested in. This 
seems like a very powerful idea: not necessarily this exact formula, 
but the idea of inducing members of the general public to form 
promotional "guerrilla armies". By way of illustration, I almost 
never click banner ads, but i almost always follow links sent to me 
by people I trust - this explains unforseeable meme-breakouts like 
'all your base' or that 'i kiss you' guy.  There is also an analogy 
here with the "relevance" engine employed by Google - counting the 
number of links to a given document as a measure of importance. This 
is not machine intelligence, but aggregated *human* intelligence.

A functioning 'fairshare' system would provide ratings skewed in 
favour of obscure or under-exposed artists, providing in effect an 
index of artists whom fans believe will be *seven times more popular* 
than they are now.  Tying a payment system in with a promotional 
system also achieves some insulation from "spamming" - if ratings are 
dependant on actual payments, they will be somewhat less susceptible 
to the types of abuses that make on-line polling so wildly 
unreliable. Combine this with a collaborative-filtering algorithm, in 
which ratings are only visible in the degree to which they intersect 
with one's own - I'm not interested in the opinions of people with 
lousy taste ;)

The RIAA-backed subscription schemes will simply reconstitute the 
current strong-arming for shelf-space and airplay on the net - with 
the same dubious personages deciding what we will or won't hear. It 
remains to be seen if anyone is the slightest bit interested in this 

A promo/CF system which is not subject to manipulation isn't a nicety 
but a necessity. It's not too late to realize the potential of 
digital networks to revolutionize the way music is discovered and 
remunerated, but it will take a serious effort of cooperation between 
artists, fans, and hackers to pull it off.  Keying the rating system 
to payments rather than file transfers should make it relatively 
immune from litigation, and make it adaptable to any distribution 
system the user chooses - limewire, tower records, or even MusicNet.

Jim Carrico

-- [Feb 2005]

your Amazon recommendations - Jahsonic - early adopter products