> On the other hand, it’s missing what I think is a major problem. Much of academic systems work already seems bottlenecked on selecting which problems to pursue. This appears to be two reasons. The short-term reason is disconnection from the problems that the customers of systems (in industry and the wider world) face. The longer-term, and more important, reason is that coming up with a vision for the future is just much harder than hill climbing. It takes more experience, more insight, and more vision to choose problems than to optimize on them. It takes more taste to reject noise, and avoid following dead ends, than to follow the trend.
While this is certainly inarguably true, I think the whole point is that AI-Driven Research for Systems as the authors put it, makes this much less critical. The sheer volume of problems we'll be able to solve will drastically minimise selection paralysis.