Comment by hdivider | Hacker News Reader

Comment by hdivider | Hacker News Reader

Reminds me of how different DARPA was in the early years, compared to now.

If you're interested in how the modern Agency functions, I'd suggest reading The DARPA Model for Transformative Technologies.

It describes some of the typical trends in our time: contracts too often go to entrenched players and primes, rather than startups as primes due to technical merit rather than past performance and relationships. DARPA personnel reaching out from anonymous mailboxes post-RFI submission, asking if your team has TS/Sci clearance. (No answer if it's not a complete 100% yes.) And landlordism: extracting value from an existing system rather than seeding many new technologies.

However, there is still nothing like DARPA, despite its many problems. DIU is more for near-term innovation, and had its budget cut 20% just now. AFRL, AFWERX, Army SBIR/STTR, and so on, are all critical but not built to prevent technological surprise. Thank your lucky stars we have DARPA still up and running after so many decades.

To me the biggest sign that this is from a different era is seeing a Director personally generate something of value! It seems like leadership are all figureheads or apparatchiks or MBAs these days, usually subtracting from the mission rather than adding.

The cult of professional administration is the single most destructive thing in America today, in my opinion.

What is the cult of professional administration?

Not GP, but for me the phrase is referring to the idea that everything needs to be professionally administered like an enterprise business (e.g. government, colleges, etc.) with organizational structures that reflect that. It typically values process and procedure over curiosity and innovation.

This is a cultural plague in pretty much all activity these days. Even in the age of memes, influencers, and user-generated content, everything has been professionalized,

Professional sports has overwhelmed real amateur leagues. The Music Industry has displaced a lot of just getting together and singing badly. Wedding planners, summer gap year planners, even professional Christmas tree decorators. There are few roadside restaurants that don’t basically reheat something delivered frozen by CISCO. Random roadside motels and amateurish amusement park attractions have vanished in the face of chains. Even tinkering happens at a high level.

Sure, all the non-professional things still exist, but are a miniscule fraction of how they used to be.

> delivered frozen by CISCO.

I think you mean Sysco. CISCO sell routers, Sysco sells food

Yes, brain fart. Thanks

I was more talking about the cultists themselves, and what sort of person becomes a professional administrator (hint: very few have ever actually directly built or accomplished anything more complicated than assembling a piece of IKEA furniture, and most of them even have to subcontract that). The collective noun for cultists is cult, hence my words.

Their collective delusion is that a business or other effort can be administered without actually understanding what it is and does. This is lethal to the business.

That said, an actually good, down-to-earth professional administrator is worth their weight in platinum. These types are very, very rare.

I have worked with ERC and many other funders, as well as in DARPA-funded projects. Talked to principal investigators, attended panel reviews, seen how projects are evaluated, etc.

My impression is that, despite some problems, DARPA is outstanding compared to what we have across the pond. Obviously, that is not a reason to ignore issues or try to fix them.

In fact, new initiatives are often trying to mimic DARPA. For example, Wellcome LEAP: https://wellcomeleap.org/leap-innovation-model/

Yes -- I hear encouraging things about ARIA. It hasn't been stood up yet, but Peter Highnam, former DARPA deputy director, was supposed to be the first ARIA director, if I recall correctly. I met Peter very briefly, years ago. Good guy.

I think Grand/Urban Challenge was the last great project and success. The openness for participation was unprecedented resulting in lot of different startups taking part (from a "serious" point of view for example that Lewandowski autonomous bike was in the GC 2004 basically an utter laughing stock of a failure unimaginable in any "serious" setting, yet it was there too, and happened to be one of the key starting points of the modern autonomous industry). I think that success kind of scared the "serious" system so that even for Urban Challenge they already had a parallel "crony" track, and seems there haven't been such an open project since then.

Wrt. original post - interesting how the catechism is different from the today's dominating approach of "we don't really know what we want, just have a very vague set of desires, lets try to achieve that small goal for that sprint which hopefully will move us in some desired direction, and we'll see after the sprint where it would lead us, and where we can make the next step from there."

DARPA PPAML (probabilistic programming for ML) was pretty good.

It helped Stan reach maturity, and also led to things like Pyro.

But we need to wait a bit more to see where this goes. The program only finished 5 years ago.

Reply on news.ycombinator.com