[1] https://www.reddit.com/r/slatestarcodex/comments/cqxcey/how_... (16 comments)
[2] https://twitter.com/alexeyguzey/status/1162017187919478784 (28 responses)
An additional trend I've noticed is that many senior researchers tend to veer away from the pure research and discovery that marked early triumphs. And veer into issues of cruft, turf wars, plumbing, and "technical debt" later in their careers. A kind of "low hanging fruit" mentality sets in. In ML research for example, issues of reproducibility and generalizability need to be solved before trust can be placed in the system. And while I recognize the ethical need for theoretical foundations to be rock solid before deployment at scale. I am wondering if that's the class of problems the most brilliant minds should be working on.
It's only now that we are getting hard metrics about the return on research expenditures. And while its currently in vogue for administrators to use these stats competitively and make claims like "UCL published more Neuroscience papers than MIT last year". We should start to see more wise allocations of research dollars into the areas of development most needed.
Thanks!
Wouldn't want to point to specific areas - rather I think that we need to find brilliant people and back whatever they're working on.
I've made a pretty big discovery, which is on its way to Nature and bioRxiv in a couple of weeks, so I think it's a very fertile field compared to pure CS (where I was before).
But working in Life Sciences is also a much worse experience due to the big amount of politics and naivety your article discusses so well.
So I have serious doubts as to whether I should continue my career in academia, despite great offers, pursue my ideas in a startup, or generate some patents and move back to CS.
Personally, I think that many of these points are also valid for computer science. However, for CS, I think short-terminism (which the essay mentions as less of a problem than initially thought, although it features heavily for example in point 9) would be well worth investigating as an important additional issue, because building software systems can take so long.
Many interesting and worthwhile long-term CS projects (such as complex compilers for new languages, new operating systems, intelligent teleteaching environments etc.) had to be terminated or significantly changed or restricted, or cannot even be started in an academic context due to this reason.
PS: have you thought about writing these findings up to publish? I'm not exactly sure of the venue, but maybe even a slightly more general audience magazine?