DARPA is the US Ministry of Planning.
The people in America who actually know WTF they are doing give it a nice militaristic name to get the twits in Congress to keep funding it, but let's not be stupid about this. Its job is the same as any other ministry of planning, in Russia, China, or Japan --- to look into the future, make guesses as to where technology is going, and try to steer research in such a way as to benefit (for some meaning of "benefit") the US (for some meaning of "US").
Where the US has a great advantage over other countries (including China) is that the totalizing tendency has not got very far here. So while DARPA is out there doing its thing and supporting its favorites, we also have NSF supporting a different set of favorites. Or ONR supporting yet another set. Or the National Labs doing their own thing. Or NASA supporting yet another approach. Or various SBR programs supporting small companies. Or even research funded by the various large foundations.
It's easy to claim that this results in higher overhead, duplication of effort, blah blah. But we have more than 60 years of experience with the US system, along with a variety of other national systems, and, really, there's no contest. The very minor additional overhead in the US system is vastly compensated for by the fact that malicious or simply stupid or shortsighted individuals cannot capture the system as a whole. The head of DARPA may, for whatever reason, be absolutely unwilling to spend money on, I don't know, THz lasers --- maybe he thinks they'll never work, maybe he hates the guy who invented them, maybe he thinks they're so "obviously" desirable that private industry will invent them without government money. Regardless, if UCLA wants to work on THz lasers, they're not screwed; they can still talk to ONR or NSF or a dozen other funding sources and IF they have a credible story, they'll probably find someone willing to help them.