slatestarscratchpad Okay so since su3 is gone and…


Okay, so since su3 is gone and we can talk about this politely now – what is going on with MIRI’s research output?

Explanations I won’t accept – MIRI isn’t hardworking (I’ve seen them, they are), MIRI isn’t smart or well-credentialed enough (they’ve hired a lot of math PhDs, former math professors, etc, all of whom have written good papers on their original subjects), MIRI is a scam (come on…)

Explanations that seem plausible:

1. MIRI are among the only people addressing a complicated problem and aren’t sure where to start.

2. People who said their research output was unusually low were comparing apples to oranges in some way or another.

3. MIRI feels like publishing their results is a waste of their time, they’re happy enough to just have them and share them privately with the relevant people.

4. There’s been some kind of delay in getting publications but it will start happening any time now.

5. Everyone else optimizes for publish-or-perish least-publishable-unit type of things and MIRI doesn’t do that, so they unfairly look like slackers.

Note that this isn’t an attempt to justify/apologize for MIRI or an attempt to attack it, it’s me trying to figure out something that doesn’t seem to fit given what I know.

@nothingismere, @yudkowsky, any thoughts? If this conversation gets interesting enough I’ll try to get info from Nate too.

I feel that the comparison of MIRI against a university research group is not appropriate.

Since the su3 debacle is making it clear that the rationalsphere cares deeply about credentialism, I’ll note that during my PhD studies I produced 3 first-author journal papers, 7 first-author conference papers and 14 nth-author papers. I was an unusually productive student in my department. I was in a practical engineering field, not CS or math.

And my God were those papers a waste of my resources.

University research groups operate within a truly terrible incentives framework. The imperative to publish X papers of Y “significance” per Z years squanders resources that could and should be used to push the work forward.

If I could do my graduate school career over, and if I were optimizing for actually-making-progress rather than appearing-super-productive, then I would refuse to publish more than maybe three or four papers (of any kind) overall. Papers and conferences are a TREMENDOUS timesink. I at least harbor a hope that the people at MIRI know this. And even if they don’t think in those terms, the fact remains that they aren’t trapped in the awful incentives framework that prizes number of publications and IMPACT FACTOR.