Why and how does DARPA work?

Program Managers

At the end of the day the ARPA Model depends on badass program managers. Why is this the case? PMs need to think for themselves and go up and down the ladder of abstraction in an unstructured environment. On top of that they need to be effective communicators and coordinators because so much of their jobs is building networks. There’s a pattern that the abstract qualities that make “great talent” in different high-variance industries boils down to the ability to successfully make things happen under a lot of uncertainty. Given that pattern, the people who would make good DARPA PMs would also make good hedge fund analysts, first employees at startups, etc. so digging into people’s motivations for becoming a PM is important. More precise details about what makes a PM good prevent you from going after the exact same people as every other high-variance industry. When ‘talent’ isn’t code for ‘specialized training’ it means the role or industry has not been systematized. Therefore, despite all the talk here and elsewhere about ‘the ARPA Model’ we must keep in mind that we may be attributing more structure to the process than actually exists.

DARPA program managers pull control and risk away from both researchers and directors. PMs pull control away from directors by having only one official checkpoint before launching programs and pull control away from performers through their ability to move money around quickly. PMs design programs to be high-risk aggregations of lower-risk projects. Only 5–10 out of every 100 programs successfully produce transformative research, while only 10% of projects are terminated early. Shifting the risk from the performers to the program managers enables DARPA to tackle systemic problems where other models cannot.

That is one excerpt from a new and excellent essay by Benjamin Reinhardt, one of the best pieces of this year, via Patrick Collison.

Note also that DARPA underpays staff, does not hire individuals with a significant web presence, deliberately stays small, and makes it easy to reallocate funds on the fly.  The program managers do not work there for any longer than four or five years, by design.

Comments

This sounds like common sense, based on my life in engineering.

Paying workers with taxpayer dollars allowing no profit with results available to everyone by law is clearly far superior to for profit private sector which promotes keeping secrets to hinder innovation.

As an engineer, life was best when we got paid to work, free to spread around what we found, especially to customers, creating more demand, and good will, for our products which everyone on Wall Street argued was the best technology everyone knew about, but bad marketing, code for a lack of profit maximization and rent seeking, just rapid growth in production and market growth with market share growing slower. That model was defined as bankruptcy because my employer only had billions in cash, billions in assets, both real estate and factories, the best technology, but not the profits of Microsoft, nor the sales growth of Compaq (which was losing money on sales at prices cut to win business).

It's a good thing my employer was taken over by a firm taken over by another, selling off assets to firms that tried rent seeking and monopoly profits, who are now mostly lost to history. Not even Intel has survived the industry MBA engineered establishment of its complete monopoly on chip technology.

The US economist and MBA model of progress totally and absolutely rejects the DARPS model of paying workers to work building new exciting world changing things.

World changing means monopoly get made obsolete and that destroys wealth based on monopoly.

nope. its about demographics. 75% of the rich world are sheep who do as little as possible to keep their jobs but are afraid of losing their daily routine. 5% are wolves who conquer and crush and aspire. They both hate each other and would collapse if either number increased or decreased significantly in the same way as eco-systems collapse when predators or prey expand or collapse too much. Therefore, there is no 'way' that is superior only the delicate dance of the ecosystem. It is this 'positive' conflict that creates productivity and wealth. The other 20%, by the way, are the beavers - non-religious/ non-political, over-achievers, who are slightly underpaid but get the job done and are responsible for most STEM behind-the-scenes success. For-proft, non-profit, NGO, small, big, union, blue collar, white collar, management, government -- all different variations on the sheep-wolf dynamic, with some Bs sprinkled in to keep us above (and below in some cultural/sociological measures) all other rich countries.

Respond

Add Comment

Respond

Add Comment

Disagree. It's about the ecosystem and timing. Case in point: Elon Musk isn't a god - he is the lucky trust fund kid who excited the right staff and colleagues, then the right investors at the right moment in time that his ideas would gain traction as is 'followers' continued to wow everyone. Steve Jobs was it. Gates has it. Ventner has it. That Virgin Airlines/ Galactic guy has it. 30% person. 70% circumstances. Amazing that such people would work for pseudo-govt, but there it is.

Was this for me? I'm not sure it aligns with the quote above, or my endorsement.

"PMs need to think for themselves and go up and down the ladder of abstraction in an unstructured environment."

In dynamic companies people at all levels jump on problems and solve them.

I know someone who shifted from California to Alaska, working in petroleum engineering. She was shocked. She learned that she shouldn't casually ask for project estimates in Alaska, because the answer was often "done."

She actually asked how hard something would be, and two guys flew out in a bush plane and did it. In winter.

Respond

Add Comment

Respond

Add Comment

-5

You’re not an engineer, it’s obvious from the remedial math incompetence. There’s a 0% chance you ever took anything beyond basic Calc.

LOL, walking definition of "butthurt loser"

Respond

Add Comment

Respond

Add Comment

Respond

Add Comment

'Why and how does DARPA work?'

To keep the American defense industry in a position to make a lot of money in domestic and international markets, using large amounts of taxpayer money.

But they don't actually make that much money anymore.

Other nations sell lots more in effective weapons used by thousands of entities with not that much money.

And the US government pays American workers half the money workers were paid in the old days of radical leftist socialism, ie, the 40s, 50s, 60s, with conservatives slashing government spending starting in the 70s. Not even massive conservative deficit spending has taken America back to the defense industry glory days of leftist big government socialism.

Winning wars quickly costs too much in blood and treasure,, followed by drastic spending cuts, so the solution that maximizes spending losing slowly and never addressing the root cause, the lack of an economy that rewards building capital instead of destroying capital.

DARPA's most popular results have been useful in defying the capital to build, but that capital does not get produced by the war industry. Eg, Internet, GPS, spaceflight, advanced manufacturing, advanced materials like titanium and especially plastics, have not done much for the war industries.

For the DARPA withheld from the public at least, it's been mostly bloat once it leaves DARPA. eg, linear motors, aka rail guns, has been applied to aircraft carriers, resulting in the most advanced carrier that can't launch anything but helicopters, if they can be gotten to the deck.

What saddens me is leftist latched onto conservative arguments for cutting spending so jobs and job training have been killed along with the destruction of a military and infrastructure for solving global problems by paying millions of American workers.

Invention and innovation can't eliminate the requirement to pay workers to build capital.

Respond

Add Comment

Respond

Add Comment

"does not hire individuals with a significant web presence"

My personal knowledge of security clearances is receding into the past, but higher level clearances rarely were associated with people in any public light, especially if it involved self-promotion.

One assumes that being part of DARPA involves high security clearances.

No

Respond

Add Comment

Respond

Add Comment

Tangentially related, because the one of the UK govt's new aims is to establish something modeled on DARPA: This recent speech by Michael Gove "The Privilege of Civil Service" has been getting a lot of attention: https://www.gov.uk/government/publications/the-privilege-of-public-service-given-as-the-ditchley-annual-lecture

Opens with Gramsci, closes with FDR.
Many interesting "average is over" points, though credited to later UK writers rather than our TC. Big on many of the MR tropes, in fact.

" There of course also a particular merit also in investing in the literal experimentation of pure science. As the success of DARPA in the US shows, sometimes by design, and sometimes by obliquity, hugely beneficial innovation can occur. Of course, some of the projects in which DARPA has invested have failed and flounderd but the knowledge that high ambition is supported and incentivised and wrong turnings accepted as necessary costs along the way has brought huge benefits.

Sadly but far too often, innovation in Government is treated as though it were a mischief rather than a model. The default mechanism of the NAO, PAC, other select committees and various commentators is that any departure from the status quo must be assumed to be more downside than upside. Had they been able to interrogate George Washington, Thomas Jefferson and Alexander Hamilton in 1783 they would have concluded that American independence was an expensive, untried and unjustifiable innovation. In Treasury terms they would have said it would have been novel and contentious and therefore had to be stopped. "

The problem is that when your own fellow ministers are meeting with billionaires bidding for government largesse or seeking favourable planning decisions, it becomes hard to ask the National Audit Office to trust in a pattern of brave risk-taking for what may well be preferential treatment.

Respond

Add Comment

Respond

Add Comment

"Note also that DARPA underpays staff........ The program managers do not work there for any longer than four or five years"

I bet they make it up in their next job. How'd you like to have DARPA on your resume? And when asked about your work - "It's classified, but you'll be hearing about it in a couple of years"

The people hiring you already know about your work. They probably knew you before your DARPA stint, and you may have funded them as a PM. Another revolving door?

Respond

Add Comment

Respond

Add Comment

"Attention science fiction fans, mad scientists, and high-tech companies with radical, life-changing ideas: the Defense Advanced Research Projects Agency wants to hire you. Every year DARPA sponsors numerous projects aimed at dramatically advancing the capabilities of the U.S. military. But what makes DARPA stand out from other government agencies seeking the same is simply the “crazy” factor. DARPA doesn’t want to hear your ideas for a remote-controlled helicopter; that’s so 20 years ago. DARPA wants to know if you can create remote-controlled cyborg insects that can fly reconnaissance missions, or produce a submersible aircraft that can maneuver under water. Or how about a shape-shifting robot that can slide under doors? All of these are real DARPA projects. If you have an idea that pushes the “crazy” envelope and can bring the future to the present-day military, DARPA has a job for you."

DARPA's history is about as strange as the projects it supports. In 1990, the Democratic controlled Congress expanded DARPA's funding and scope to permit investment by the agency in the high-tech companies which were related to the military. That was beyond the pale according to the Bush administration, which sought to curb the co-ownership option: it was okay to provide financing but co-ownership was too much like state capitalism (the kind that would become familiar in China). It's not surprising then that PM turnover at DARPA is high. Should DARPA, should the government generally, end the pretense and invest directly in the companies with which it does business?

Those familiar with the China model know that with the symbiotic relationship between government and business comes overlapping governance, with government (political appointees) working together with company executives sharing management responsibilities. I'm not trying to change the subject, but Cowen may be saying more than he lets on with this blog post.

Respond

Add Comment

My last comment got lost in the ether. I was saying DARPA only lists 12 patents or patent applications as assignee according to Google Patents, which means either they keep stuff trade secret or they fund other entities that go on to patent inventions. Their Wikipedia listed big projects sound cool (https://en.wikipedia.org/wiki/DARPA#Projects) but a bit too sci-fi, though I like the "insect allies" projects (using GMO and cyber modified insects to conduct warfare).

Bonus trivia: Wuhan Institute of Virology (WIV) vs DARPA - who has a bigger budget? DARPA ($3.4 B /yr) Who has had a bigger impact on the world? Snicker, snicker, it's not even close...

Bayh-Dole.

Respond

Add Comment

DARPA generally expects the “traditional” companies awarded contracts to buy in, either by charging the contract at no-fee rates (i.e., overhead only, no profit), investing company money (cost share), or both. The traditional companies are also expected to bring non-traditional suppliers such as universities, small/minority-owned/woman-owned businesses, etc., onboard with significant statement of work. To balance this risk, the contractor usually retains significant intellectual property rights to the work done. So any patents generally accrue to the contractor, and the government may get preferential terms for licensing, but the government seldom owns any patents developed on the contract.

It’s not always this way, but it commonly is.

Respond

Add Comment

Respond

Add Comment

Respond

Add Comment

Those look like useful thoughts, although I'm wary of what's basically a post hoc explanation, sort of like how I can come up with great explanations for why the stock market will go up or down today -- as long as I'm doing the explaining tomorrow instead of yesterday.

In recent years I've gradually become more aware of the importance of program managers. That doesn't mean that Reinhardt is right, but he could be onto something.

So this essay is worth pondering. I've been unimpressed with at least 3/4 of the "progress studies" links that Tyler has posted here on MR; they've had a lot of thoughts that haven't been completely thought through or that are ignorant of basic or at least countervailing facts. But I suppose we could say that's actually a good performance, based on Sturgeon's Law that 90% of everything is junk.

Respond

Add Comment

The incentive for PMs is 20% extrinsic--you will know the whole waterfront of a field and where its going by the time you leave, which can be monetized/operationalized in many ways--and 80% intrinsic--the government will give you $60 million to do the brilliant things you always wanted to do in technologies that you have been pursuing for years.

Respond

Add Comment

Excellent description if a little bit like someone describing secondhand what makes sex amazing. Would have liked to hear more from principal actors. Something magical happens on projects that push limits and succeed against great odds.  The most amazing thing of all is how some PMs are able to keep replicating their success like Licklider at ARPA and Satya Nadella at Microsoft.  I suspect Licklider honed his instincts and amazing people networking skills at BBN. Talk about a group of star performers. Got a kick out of seeing TC's production function question. I suspect the best PMs have a talent for figuring out what turns people on and how to get them firing on all cylinders. 

Respond

Add Comment

DARPA is the closest thing the US has to industrial policy. They typically choose a promising area based on discussions with researchers and then set up an "initiative" that might run for decades. Initiatives are driven by project managers who are usually chosen from among the various research groups for a fixed term, typically three years, but sometimes longer. They area almost always scientists or engineers, not MBAs. They are given funds to spend and usually try to back multi-year projects at a number of institutions with some record of research in the area, but anyone can reply to an RFP.

The advantages to this approach are that they have a long attention span and they can offer long term funding. Project managers are given a lot of freedom. Since there is a long time horizon, the definition of project success might include "failure". Even failures provide useful information and educate people working in the project area. A DARPA failure can lead to a private sector success.

I was involved in a parallel computing initiative. Our group focused on artificial intelligence applications. One application, among many, was autonomous vehicles. This was back in 1986 and our PM was almost rammed by the minivan CMU had rigged as a test bed. (That would have been a serious failure.) It would be another TWENTY years before this technology started interesting venture capitalists, though NASA used a lot of its technologies in its Mars rovers and other spacecraft. By the early 1990s, the same DARPA initiative, probably with a new name and new PMs, was funding things like workstation computers with high enough bus speeds to drive a 1080p HDTV display, something an Apple Watch can do today. I retired in the late 90s, so I've only been following DARPA through the media since.

There have been all sorts of other initiatives in energy, biology, chemistry, manufacturing and so on. They all have loose, sometimes very loose, military justifications e.g. bullets are made out of matter, and this project involves manipulating matter. Unlike China, Germany, France and other countries, industrial policy has to sneak in through the back door here in the US, but DARPA shows us what is possible when one lets the government pick the winners and losers.

Respond

Add Comment

"Note also that DARPA underpays staff"

If you are a libertarian does it make sense to say that someone is underpaid? From a libertarian perspective people earn what their output is worth in the market. And the market Outcomes have no ethical significance

Maybe the more nuanced verbiage is "DARPA underpays staff salary-wise in exchange for non-monetary benefits/prestige"

Respond

Add Comment

Respond

Add Comment

So DARPA is sucking up all the competent PMs in the United States while the rest of us have to sit through our fifth straight "Status Meeting" where nobody has completed any of their action items?

Great :-\

Respond

Add Comment

I have read your Article it is knowledgable and Informative.
I appreciate it.
ARPA Model is a Good one. and according to me, DARPA is not Good Enough regarding Work and payment parameters.

Respond

Add Comment

Respond

Add Comment