(I'm going to agree with your first sentence by disagreeing with everything you said about market making)
> Look at market making (providing liquidity for other traders). There's an intrinsic demand for liquidity from funds which try to manage an index, hedgers, and speculators. The goal of market making is to provide liquidity to facilitate those other traders. And yes, there's a small premium for the service, but the fact that the premium is shrinking over time reflects a lower cost for the rest of society.
I think that's a very quaint view of market makers, maybe valid 20 years ago but it seems disconnected from modern circumstances. Market making has largely become synonymous with algorithmic trading, which now accounts for somewhere between 25%-75% of traded volume (obviously varying widely between markets, see http://ftalphaville.ft.com/blog/2011/03/04/505021/algo-tradi...).
A significant subset of the prop traders have been spiraling into a latency arms race, which has gotten to this absurd point where even millisecond-scale trading is considered slow.
Regular old market making (put orders on both sides of the book, make profits from the spread) only seems to work in the absence of significant competition. In reality, profitable firms are running all sorts of short-term speculations and even trying to prey on the trading behavior of people & other algorithms. There have been all sorts of algo-induced pricing anomalies documented at http://www.nanex.net/FlashCrash/FlashCrashAnalysis.html.
The supposed benefit of all this nuttiness is improved liquidity, but I wonder: Who needs liquidity at the millisecond scale? What business, other than algorithmic speculation, will suffer if they have to wait 1/2 a second to buy or sell something?
Also, I question whether the premium of modern prop trading is actually small-- there are a large number of prop trading firms that seem to be doing extremely well for themselves. I don't know the specifics, but I wouldn't be surprised if the sum of prop trading profits reach high into the billions.
Furthermore, these companies siphon off a large number of smart grad students, who would have likely otherwise done something actually socially productive with their time.