I definitely think 20% is not that crazy, especially considering the gains those railroad investments are probably still paying to this very moment.

I can give you some anecdata on this. Just yesterday I was talking with my Trump-supporter family members, and they literally had no idea he'd been convicted for felony fraud - they thought his only charges were related to the sexual assault accusations from E. Jean Carroll and they hand-waved those away with gusto anyway.

So to answer your question: It's all fine and going on as expected for them, and if you tell them otherwise you're labeled an alarmist crybaby democrat.

I think you're joking but there was a company in my area that famously stuck by their "mob programming" approach where 5+ developers would sit in a room, 1 would drive, and the rest would fill out an adjoining document.

We’re so good at it we’re going to outcompete ourselves in fact

Replication is great but it’s not the only way science learns and it would be foolish to put that kind of limit on publishing your work.

That seems appealing but then as the applicant you’re at the whim of a LLMs biases, which are even harder to discern than humans.

Except that’s a regular experience I’ve encountered while job hunting this winter. It’s either that or what, don’t apply to a good job that I want? The applicant has no leverage in these situations.

I got hired off of a HN thread a few years back. Worked on an open source message queue in Go for a year or so before lay offs hit. Overall, I’d do it again, HN is a solid culture fit litmus test apparently.

I think it’s more widespread than you might think. Flat earth and similar conspiracies play on distrust in academia, which is absolutely flourishing right now.

Ive seen this referred to as “AI drift” before and I see it being a long term problem of data sets for training. As always, the quality of the data set determines the quality of the results. Models and data sets will be the arms race in the LLM world for a while to come, I think.