Hacker News Reader

Comment by pbateman | Hacker News Reader

This strikes me as a much better idea than the standard process of awarding a cost-plus contract to a military contractor and then watching as costs blow out.

View on HN

DARPA's usual approach is to find some research group with a record of being brilliant and getting good results, and give them enough money and an ambitious-but-probably-possible goal. They're a surprisingly efficient government agency.

I'd reserve judgment until something actually comes out of it... More generally, "prize funding" isn't a practical mechanism for building real defense projects. Nobody who knows what they are doing will incur huge capital costs for a mere chance at some prize money.

The 2005 DARPA Grand Challenge [1], a self-driving-car competition, was extremely successful. Winning entrants spent far more than the prize of $2 million - predominantly funded by sponsorship from companies interested in self-driving vehicle technology.

[1] http://en.wikipedia.org/wiki/DARPA_Grand_Challenge

Self driving cars is far more sexy project than this. The PR that universities got from that will be way way more than for this.

Not a great example. A lot of the work on DARPA grand challenge was done by universities, which as a group are heavily funded by the federal government.

As I understand it most university projects don't get much cash from the university itself. In general STEM research is funded by grant money/sponsorship from industry that academics raise themselves. Often it's the academic/research group that then pays the university!

Admittedly the two leading teams did have a bunch of undergraduate students working on them, which offsets their labor costs - and the univesity usually supplies some resources.

Tell that to Sebastian Thrun (http://en.wikipedia.org/wiki/Sebastian_Thrun), whose work on the Google driverless car project (http://en.wikipedia.org/wiki/Google_driverless_car) and current position as a Google Fellow owes so much to his successful work on previous DARPA challenges (http://en.wikipedia.org/wiki/DARPA_Grand_Challenge).

Granted, the challenges cited above didn't directly provide any market-ready products, but the practically-minded research and development they induced significantly advanced the algorithmic and engineering know-how available to those looking to create marketable artifacts. More generally, reputation is hugely important among researchers and head-to-head competition is a cost-effective way to generate and distribute reputation.

It is almost certainly the case that Sebastian Thrun's research was paid for by the US government in the form of research grants to academic institutions, and student tuition.

So yes, prize money works great if you're already paying for research, and you just want to prod the direction of that research a bit.

On the contrary, it has an excellent historical record. That's how we wound up with accurate clocks (for use at sea a couple of centuries ago).

For anybody who wants to know more about this:

http://en.wikipedia.org/wiki/Longitude_Prize

The capital investment for inventions a couple of centuries ago is totally different than today. We live in the age of Big Science, where scientific advancement requires large teams of highly skilled people using very expensive equipment and instrumentation. A lot of the advancements in physics in the late 19th century and early 20th century were done by one or two researchers with small scale university resources. Today, these same sorts of advancements are done by international teams of scientists using particle accelerators that require massive government investment by multiple countries.

I don't agree. Past innovation looks like low-hanging fruit, but consider things like the breakthrough production of graphene using nothing more elaborate than scotch tape. given the advances in computing power, you can get a lot of good work with a potential reward of $100k or $1 million.

I think you're right and wrong.

Coming up with awesome new SDR waveforms is something someone who is already funded can do, or something someone can do for fun (I'll probably enter). In this contest, the cost to enter is really low, and you'll get radio hackers who are interested in the problem. DARPA wins if it gets a few marginal people interested in the problem, not just if they come up with a new production waveform.

Contests work well if they're easy to enter or if winning brings you notoriety beyond the prize. Otherwise, you'd have to set an absurdly high prize to make it a high enough expectation to facilitate raising money for each of the teams, enough to have real competition -- like offering $1t for 51% of Osama Bin Laden in 2001, or $20t for a successful Mars colony for 100 users, etc.

Reply on news.ycombinator.com