Michael O'Hare's Fuzzy Math

So what is possible? You could easily scale the problem down; start by recording only studio produced albums (i.e. exclude all concerts and radio broadcasts, but allow small indie artists to cut a few tracks and put them up even if they're not on a label). That means you're looking at a fixed annual cost of around $2 billion ($4 billion if you include all movies), which is what we spent on constructing the superconducting supercollider that we never used, only in this case we're spending that much money every year. That's a manageable number, but I would still prefer that my tax dollars go towards some massive Apollo Energy Initiative focused on conservation, energy efficiency, renewable fuels research, and carbon sequestration, rather than to building some sort of nationalized iTunes. We can let the private sector sort that out, even if it means putting some money in the pockets of the RIAA.

But is it even a good idea? I'm not sure. The formula used to calculate royalties would be subject to lobbying pressures. There's the user fee argument that there are people who don't listen to music or watch movies and therefore shouldn't have to pay for other people's music habits (and we've left the realm of the NEA and its $20 million dollar budget; this is Real Money). I should say that the State probably can have some greater role in helping monetize the long tail and encourage artists to make a living at a level in between "I wait tables to pay the rent" and "I'm a Ca$h Money Millionaire" and thereby encourage more people who might not otherwise make music or movies to make music or movies. But I think if you support this sort of thing you have to rethink your opposition to farm subsidies.

It adds up to real money faster than you think: This exercise should illustrate that creating some new government program whose cost is measured in billions is harder than you think. Because the national political debate is focused on Taxes, Social Security, and Health Care, and both attack ads and positive ads usually use some multi-year estimate to make an even bigger splash, we tend to think that finding a few billion dollars in the federal budget is not that hard. But the non-defense, non-DHS budget funds a huge number of agencies, and really only amounts to $400 billion a year. Squeezing in the COPS program, with an annual cost of $2.6B a year, requires moving around a substantial chunk of the overall federal budget, and represents more than 10% of the Justice Department's budget.

So, be thinking of all the programs you can cut to make room for that Apollo Energy Initiative.

UC-Berkeley professor Michael O'Hare wrote an op-ed, which got this whole game started, and now he's got both Mark Kleiman and Matt Yglesias extolling the virtues of some state-run archive of all music, books, or movies ever made, that provides the citizenry with on-demand access to the entire catalog, with the government doling out money to artists based on the number of times their work is heard. Before this gets out of hand, I feel the need to burst some balloons. In short, producing a system capable of delivering more content than our existing television/cable/radio infrastructure is an immense problem. Professor O'Hare needs to go across campus and ask David Patterson if he (qua Patterson) thinks his (qua O'Hare) National Media Center idea is even remotely feasible.

The amount of audio and video media the US produces is completely dominated by television, because there are so many stations and they produce new content with great frequency. One year's worth of television programming stored at DVD quality would consume between 4 and 12 petabytes, or 12 million gigabytes, of disk; for comparison, that's equal to the disk storage of 200,000 60 GB iPods or ten times the amount of disk space used by all computers at Amazon.com. Radio broadcasts would consume a similar amount of disk (they consume far less space, but there are many more stations). Every year, then, it would costs $2.5 billion dollars just to store the media produced that year at current enterprise storage costs. So if we started taping tomorrow, by 2030 we would be spending $62 billion in today's dollars just on storage under the "high cost" estimate ($25 billion under the 4 PB/year "low cost" estimate). By comparison, the budget for all of NASA was $16.4 billion last year. Sure, some technology breakthrough might cut costs in the mean time, but a different breakthrough might also cause storage demand to explode.

The even bigger challenge is piping all this data out to the masses. Let's assume you'd need about 100 times as many "channels" as we have today to satisfy everyone's viewing habits at any one point in time. Sending that much data would require a network capable of delivering 1.5 terabits per second, or roughly the 1 million times as fast as a very high speed cable modem connection. The world record for sustained network over a long distance, under very controlled experimental conditions, is 101 gigabits per second, or one tenth the required speed. At this rate, in six to eight years some researchers might be able to get close to our bandwidth requirements. Even if I am off by a factor of ten, it will take decades before this sort of network existed under commercial conditions. And it would require a huge capital investment to build.

Update: The Professor himself comments, saying that he doesn't care about storage and distribution; he just wants to have the government collect royalties based on the frequency that digital media are played. That's an idea I could get behind from a technical standpoint. I don't see how you solve the problems that SETI@Home and other distributing computing projects have had. What if I hack the app that's sending usage data, and claim to have played a song several hundred times even though I've done no such thing? You could make that activity illegal, but how would you tell such illegal activity from sudden surges in demand for a song? And what if I get a thousand people each to fake playing the song once, which will be much harder to detect? What if I just leave a song on repeat but don't actually listen? At first blush it seems like the opportunities for fraud are ripe.

Michael O'Hare's Fuzzy Math Now Playing: Atmosphere / God Loves Ugly / Vampires

Michael O'Hare's Fuzzy Math
Now Playing: Atmosphere / God Loves Ugly / Vampires