logoalt Hacker News

jordanbyesterday at 4:39 PM1 replyview on HN

The servile stuff was trained into them with RLHF with the trainers largely being low-wage workers in the global south. That's also where some of the other stuff like excessive em-dash stuff came from. I think it's a combination of those workers anticipating how they would be expected to respond by a first-world employer, and also explicit instructions given to them about how the robot should be trained.


Replies

OkayPhysicistyesterday at 5:33 PM

I suspect a lot of the em-dash usage also comes from transcriptions of verbal media. In the spoken word, people use the kinds of asides that elicit an em-dash a lot.

show 1 reply