Where some of the discussion about the stagnation hypothesis gets kind of confused and confusing is when you get back-and-forths that go like this: people say, we’ve had slower growth. There hasn’t been so much innovation and progress. Then other people come back and say, ‘what are you talking about? Look at all the progress in computers and the internet and communications and smart phones and now AI.’ The counterpoint to that is ‘it’s only in the world of bits, what about atoms?’
The counter to that is, ‘what do you mean if you set that aside? Isn’t that unfair? You’re carving out the area where we’ve had all the progress, then you say we haven’t had much progress. Well that’s not fair.’ The counterpoint to that is, ‘well it just one area. If you look back at previous times in history, we’ve had lots of progress in all these areas.’ Then you can argue, ‘isn’t this just how progress goes? We don’t make progress in every area at the same rate all the time. In fact, it naturally happens that an area becomes hot and there’s a lot of progress, then it plateaus, but we move on to other areas.’ And then the counter to that becomes, ‘why is there just one major area of technology where there’s a really steep s-curve right now?’
In the past, we had multiple s-curves going on at once. Maybe energy technology was revolutionize, and at the same time manufacturing and materials, and the same time transportation, and medicine, and… now we’re down to just one steep s-cruve where we used to have two or three or four going on at once. I think the way you get to clarity is by explicitly separating out different s-curves and analyzing them separately — both the individual s-curves and then analyzing them as a collection. We can say, yes, the natural course of any one technology is that it starts of slow, hits some sort of inflection point, goes into exponential growth, then eventually hits a saturation point and maybe levels off. We’ve picked the low hanging fruit, and things get a lower ROI. There’s a natural sigmoid or s-cruve where it plateaus. The thing with the low-hanging fruit analogy is that the discover entirely new orchards, there’s whole new fields of low=hanging fruit, and we can go in and that is discovering a new s-curve.
That was another excerpt from Jason Crawford on the Venture Stories podcast, “Progress Studies in 2020.” I’m just going to add that the Department of Defense used to be a leader in discovering new s-curves and then getting to the exponential growth stage. Perhaps this was only the case in the 1940s to early 1960s era. But that was the time where there was abnormally high productivity growth. But there was also a lot of invention and new s-curves at the end of the 19th and start of the 20th century, and these were basically all privately funded or through tinkering.
Today’s bureaucratic and waterfall processes perhaps make the government far less likely to be on the frontier, creating new s-curves. In fact, most of the new technologies people are excited about in defense — AI, robotics, additive manufacturing, 5G, blockchain, and even cloud — these thinks may have started somewhere in a government funded lab, but government never exploited them. The technologies had to be scaled in the commercial sector, and only by the time that the tech starts to plateau are risk-averse government officials ready to bring it into their missions.
This is a huge problem. Perhaps one of the biggest problems in defense technology policy. The DoD’s weapon systems have almost all plateaued decades ago: fighter aircraft, precision strike, ISR, nuclear weapons, submarines, aircraft carriers, ground vehicles, and so forth. The DoD will almost certainly continue stagnating unless it is in the business of creating new s-curves. The current policy of waiting for commercial tech to lead the way in new s-curves and only bring it in after it becomes common place is exactly the strategy which will put the US military behind China — and that is already happening. Without taking risk on the frontier, the DoD will lose.
In my mind, the emphasis on dual-use technology is short-sighted. Yes, these commercial technologies have great potential in the DoD. But that’s because the DoD wasn’t able to scale the tech like AI and robotics that was coming out of its own labs years ago. And then there are whole classes of systems that simply aren’t commercially viable, like hypersonics. This class of military-unique technologies cannot be neglected for a singular focus on boomeranging dual-use tech from DoD labs, into commercial firms, and then eventually (maybe) back into defense acquisition programs after it has plateaued. This is a recipe for failure.
Certainly the DoD labs are doing a lot of interesting things that could lead to new s-curves. The foundations are there. The problem is whether mainstream defense acquisition programs can grab hold of those and scale them across the valley of death. It is unlikely so long as oversight agencies demand zero risk, lifecycle plans, high TRL, and so forth, before any acquisition program can start.
Leave a Reply