Should AI companies be liable for LLM output?
(media.kotakuinaction2.win)
You're viewing a single comment thread. View all comments, or full comment thread.
Comments (48)
sorted by:
Reminder that every "survey" or "study" is almost entirely conducted on college campuses or cities near them. Meaning that the sample size is overwhelmingly favored towards young Lefties.
So these things are more of a reflection of that specific demographic instead of "US Voters" in general.
I never thought of this. But I was surveyed many times while at uni and almost never since…
a few dollars for my time was more appealing back then
The money is always good, mostly because its easy and free. And most college kids are gonna be there anyway, so why not.
But almost all of these things are operated out of such locations, so they aren't going to travel far if they don't have to and technically a campus isn't Leftist only so they can pretend its diverse.
The only alternatives you ever see are phone surveys and online polls, both of which have such a self selection bias of who the fuck would answer this that they are useless for anything other than social engineering headlines.
I recall one college Psych study where they paid me something like 15 or 25 bucks to get drunk.
God damn man, that's an absurd level of payment for one of these. We were taught to give like 5$ back in the day.
What incentive do survey companies have to ever actually ask any person any questions? They are paid to produce a result. No one asks about the historical reputation of the surveyor—we just consume the results as fact.
Just the fact that they need to do enough work that it looks like they did something, enough to satisfy their sponsor. Otherwise whoever paid them could have just paid a random PhD to attach their name to a made up quote entirely, and we aren't quite that far gone yet most of the time.
And a lot of the time these things are being done by grad students or research assistants, so there is a grade being held over their heads to at least conduct something.
You are right about the rest though, no one reads past the headline and the chart. Because if they did it would be comically obvious how weak 90% of these are in telling you anything.