Big-budget research

[markgritter]

A while back, Scott Aaronson posted his slides from a "crazy idea session" asking why theoretical physicists get 8 billion dollar machines while theoretical computer scientists tend to be funded on a less generous basis. What might theoretical computer scientists find to do with $8B in hardware? His suggestion is to look for some lower bounds on

Comments

[nsight7]

Something else I can add, as a physicist, is that in a large project like the LHC we don't get just the massive amount of data either, or just the research output of a very large number of scientists, but we also tend to increase our technological ability at each stage. Every project sees some significant advance in technology of some sort and so the windfall of the increase eventually finds its way down as well, whether it is an improvement in data acquisition technology or something that will eventually become a useful applied tool. In my instance, we borrow pretty liberally from MINOS in building our muon detector, and we are using it for cosmic-ray tomography and the applications include stuff from imaging to geological sciences to counter-terrorism and so forth. And this isn't even great stuff. The stuff we get from major projects is much better!

[markgritter]

True, although other fields have spinoffs too. Funding the HGP advanced the state of the art in genetic sequencing but also contributed to information sciences. Nearly any big engineering project will advance technology.

[prock]

I hear it only cost $12 to build the internet.

[markgritter]

We tricked the private sector into building it for us. Now they have their own ideas about what it should be used for.

It's the bombs.

[abostick59]

The reason physics gets large sums of money for big experiments is simple: the first Big Science physics projects gave us atomic bombs and radar. Somewhere along the line, policymakers bamboozled themselves into thinking that funding physics projects, as opposed to chemistry, biology, or, God save us all, the humanities, is good for national security. If it doesn't produce bombs directly it trains people who can design bombs later. (See Vannevar Bush.)

[jon_leonard]

Physics is older; CS hasn't yet come close to exhausting the things that can be learned with mere million-dollar hardware (nor, for that matter, thousand dollar hardware), so there really isn't an argument for the big toys. Yet.

With more time it should also be clearer which problems aren't vulnerable to just a bit more cleverness, and need more computer time to solve.

And finally, 'wait until next year' really does improve computer hardware (for now) in a way that doesn't apply with accelerators.