pull down to refresh
I feel this in my bones.
I just went through a 2 year publication cycle between first submission and final acceptance. 4 rounds or revision. The main results and central insights did not change the entire time.
Ahh, but were you using exactly the right kind of standard errors?
I did in fact have to redo my standard errors using bootstrapping for one of the revisions
I remember one of my grad school officemates getting super bogged down in different standard errors.
He was doing something that didn't neatly match any convention and there were about half a dozen plausibly correct methods to choose from, so he ended up spending weeks on making these big ugly tables comparing his results with each of the methods.
I don't disagree that blogging can be interesting whether it is deep research or speedy analysis. I do suspect, however, that there is a class of very popular internet personalities who are going to see their income model change.
If anyone can prompt out an llm-generated analysis of the latest thing which, despite it's tonal flaws, delivers quick, fairly interesting conclusions, what does a writer like Noah have left to distinguish himself? Maybe it becomes even more based on personality?
I'd like to say that niches are the solution for bloggers, but more likely it's that bloggers will end up writing what is interesting to themselves and having small audiences.
What I find interesting about your prediction is that the outcome may say something about whether the influencers' followings are due to the content of their thoughts, or due to the force of their personalities.
If their following was based on the words they put out, and if their words are slop-like, then yes they will be threatened by AI. if it's more of a social phenomenon, then they won't.
There's a more charitable and interesting read on the point about speed. Most of what researchers spend their time on doesn't ultimately change their conclusions.
The vast majority of the time our priors will be correct. Also, what @SimpleStacker or I refer to as the "naive regression", which doesn't account for any of the tricky empirical complications that can occur, will usually be correct.
Almost all of our research time is spent after forming our initial expectations and checking the naive regression results: i.e. it's usually effort expended after having arrived at the correct conclusion.
I've always been of the opinion that we should publicize more of these preliminary thoughts, while also acknowledging which potential issues have not been addressed that might change the conclusion.