One of our members asked me about ARPA's request for "technology transfer" plans: "a description of the results, products, transferable technology, and expected technology transfer path." That's a bit hard to document when your products are software and research papers.

Questions about proposal solicitations are best taken to the relevant program director; he or she has an interest in teaching scientists the kind of work that ARPA needs. My understanding is that they want collaboration with industry rather than just [6.1] basic research and publication of results. ARPA's survival depends on getting bang for the buck: fielded military systems from ARPA-sponsored research. The agency works with defense contractors who can move research into practice under separate 6.2 and 6.3 funding.

The US military is being pushed toward dual-use technologies and less 6.3 funding. No good method of tech transfer has ever been found other than through a customer who needs and guides your research. (Researchers themselves can transfer to industry, but that takes creative people away from research.) ARPA doesn't have a magic answer for how to transfer all technologies -- they are depending on you to solve that problem for your own particular proposal.

NSF is much less concerned with such matters because their most influential constituency is scientists themselves. An NSF proposal need only show benefit to other scientists in order to win funding. Implementation of good ideas can be delayed by many years if professors and their students are so well funded that they needn't coordinate with industry. SBIR and STTR grants are the closest thing that NSF has to ARPA-style competitions.

(I'm not sure where NIST grants fit into this spectrum, but NIST has been concerned mainly with manufacturing rather than computer science.)