Anecdotally you have stories about people putzing around on something for a while (Ideally corporate R&D enables targeted dicking around) and then roping in a few people part time, and if it shows promise going from there. In this way R&D labs seem like the only institutional structure that allows research to start small and then ramp. It is important for big things to start small.
This smooth ramp stands in contrast to academic labs where you can start small but you will hit a scope ceiling quickly outside of a few exceptions because of the size of grants and the individual publishing imperative. Academia is not a good environment for systems engineering. Grant-based funding in general seems like it makes smooth ramping hard, so national labs would seem to suffer from the same problems.
Small businesses don’t have the same ceiling as academic labs, but it is hard for them to scale smoothly, unless the work is smoothly coupled to profits. Unfortunately, some of the most interesting work leads profits, so in order to go after them a small business would need to transform itself into a high growth startup. Startups have the opposite problem, where they have a hard lower bound on size and smoothness. They’re expected to scale quickly - if a startup is not constantly hiring it’s a red flag.
Ideally, the combination of academic labs + startups should enable this smooth ramping, hopping from academia to startups. However, there is often a gap between where you can get in a lab vs. what you need to start a startup. Hence The (idea) valley of death. It’s worth spending more time thinking about the ‘dimensions’ of the valley of death - because it’s not a matter of number of people as the “ramp” language here might imply.
It’s informative to contrast Google Brain and Google X in the context of smoothly ramping projects. Google Brain feels like it is a place where there’s a ton of different things being tinkered on at once, with projects of all different scales. Modern AI R&D labs feel like corporate R&D labs of yore. Google X feels like there’s actually little tinkering before either launching into a full-scale flagship project or killing the idea. X famously prides itself on ‘killing ideas quickly’ which is a good move for startups but perhaps squanders the advantages of CR&D.
It would be interesting to plot a histogram of project sizes at different organizations. My hypothesis is that healthy ones would have a nice smooth decay - lots of little projects and a few big projects and many in between. I suspect that many organizations would have a hole in the middle. Maybe that’s The (idea) valley of death appearing as a statistical phenomena.