The idea of simply creating a different distribution mechanism (eg CDBaby) for unsigned artists so that they don't have to sign with a major label is popular with the tech crowd, but fundamentally flawed. It assumes that unsigned artists can all produce well recorded,
popular music on their own.
Most can't.
A good producer can pick a rough artist or band and get something popular out of them, but it often involves both mentoring and investment in recording and promotion upfront before any opportunity of return. Note I deliberately use the word "popular" rather than "good" (which is subjective anyway) because clearly as a business "popular" is where the money is. The "long tail" theory is fiction. Music is a hits business where most of the money is made by a relatively short fat body of popular music.
Many of the most popular "artists" are created from scratch by producers. Where do you think Britney and the Spice Girls came from? Like them or not, the revenue from hit music funds the production of new acts who haven't become profitable yet.
Another flawed assumption is that everyone wants to listen to new music. A lot don't. So you have a large body of recorded music that the record labels will never give up or let you sample or remix without a hefty toll. And as soon as any new act gets popular, you can bet they'll be offered a seat at the "big 4" music table. It's the DeBeers business model applied to music.
Personally, I think what the radio station JJJ in Australia does with their unearthed site (http://triplejunearthed.com/) is an excellent model for identifying quality artists. The next step IMHO is a ycombinator style production startup company that invests in developing the most promising acts to become popular and then offers all the recording, promotion, management and touring services that the labels do but on just terms. Distribution is just the last piece of the puzzle.