« Two Supercomputers Di… | Home | Knowing an Unknown Un… »

Even More Ways to Use the World's Fastest Supercomputer

Michael C argued last Wednesday that we could use the world’s largest supercomputer better. As he wrote:

”So, I ask you, what would be a better use of the world's largest supercomputer, trying to prevent a handful of attacks a year by terrorists or the medical innovations that could come by feeding the world's fastest supercomputer human genome data?”

Reading an early draft of the post, I responded, “That’s not the only way we could use one of the world’s greatest supercomputers.” Instead of wasting all of that technology on recording Americans phone calls and internet usage, we could use it to...

...Model Global Warming. I’m not even saying that we have to use it to support the theories of liberals. We can use it to run as many scenarios as possible. Run as much data and as many variations as you can. The government does have a supercomputer running these scenarios...and it’s only the twentieth largest supercomputer in the world. Oh, and according to research I found on Wikipedia, its memory is smaller by a factor of thousands.

...Predict the Course of Natural Disasters. The National Center for Atmospheric Research actually has a supercomputer that, in addition to modeling climate change, studies “tornadoes, hurricanes, earthquakes, droughts, and other natural disasters”, helping to improve our predictions. If it works, let’s make it bigger, instead of spying on Americans’ phone calls and internet searches.

...Model the Economy. According to Nate Silver, modeling and predicting the economy is a fool’s errand,. And he’s probably right. But imagine if this computer at Moody’s were the size of the government’s supercomputer. We’d probably get a hell of a lot closer to reality.

...Recreate the Big Bang. No life saving achievement here, but it does add to our knowledge of the world...and it would keep America on the forefront of scientific innovation.

...Figure Out the Human Body. Like Michael C described in his last post, supercomputers can help us save lives by improving medicine, folding proteins and mapping the human bloodstream in ways humans never could, two innovations that could save lives.

...Advance solar power technology. Cause, you know, that’s what the Chinese are doing with their supercomputer. Sigh.

...Something we haven’t thought of yet.

Anyway, we could do all these things. Or none of them, and spy on Americans to prevent the terrorist attacks that rarely happen.

six comments

Get off the supercomputer thing. Supercomputers aren’t scarce, they can be produced. Econ theory suggests that we have supercomputers for the worthwhile uses of supercomputers, after all. We’d build one for a given purpose unless some market failure strikes (externalities, for example).

The real trade-off is in regard to talent and funds, not supercomputers. I get your approach; you use the supercomputer as a non-abstract embodiment, but that’s where the econ foundation becomes shaky.

Sorry, I didn’t really mean “aren’t scarce”, that was an inaccuracy on my part.
I meant that using one for this purpose doesn’t mean that some other worthwhile purpose remains vacant.

@ S O – For now, we’re just doing a one to one comparison. If your point is that we should have focused on the 1.5 to 2 billion dollars (and 40 million in yearly maintenance) (assuming the press releases are forthcoming and valid) price tag, I get that.

But before we get to dollars and cents and man power, we just choose this example.

The United States has different valuations of saving a human life in different agencies. Any measure that’s pricier is not supposed to be taken in order to keep the funds for easier opportunities to save/extend lives.
This sounds cynical, but it’s the scarcity of the world that forces humans to draw a line somewhere, and you better stick to it or else you save less than you could have saved with the available budget.

Agencies which purport to save lives should publish their ‘line’, their figure (such as USD 6 million, for example: http://en.wikipedia.org/wiki/Value_of_li..). Then one can divide their budget by their limit figure and see if their activity can save enough lives to meet that limit at all.
Reality will be more complicated as the agencies will claim multiple purposes, but they could at least not get away with the claim that their work is mostly about saving lives if the calculation doesn’t support this.

This assumes that citizens or at least journalists can handle this degree of difficulty, of course.

SO – You had a post on comparing saving lives to soldier spending a while back, right?

Also, when it comes to economics, you could argue that the government fundamentally distorts the private sector by spending research funds. I would argue that super computers are still rare because of their large up front costs.


Factories have large investment costs as well, and there are plenty of them.

What drives public investments is the nature of the utility:
Does the investment (and operation) create public goods or private goods?
Is the commercial approach inferior? Natural monopolies (such as privatised, toll-collecting bridges) create inefficiencies, for example.

There’s another explanation for why supercomputers are rare:
The sum of problems to be solved by supercomputers is being addressed slice by slice. The growing performance of supercomputers makes ever more problems solvable and soon thereafter the costs of solving the problem drop very much.
I can’t illustrate this right now, but trying to improve the world by adding more supercomputers would be like trying to beat the natural GDP growth trajectory – enticing, but expensive and in the end a small advantage paid for by high sustained costs.

I would probably write a paper about theoretical supercomputer economics right now if I was in econ research! ;)