Hacker News Reader

The key unresolved question is: why do LLMs follow instructions, especially system prompt?

And I don’t see a good answer which is not worrying if not spooky.

What’s unresolved about it?

They are assembled to. We tweak the numbers, bit by bit, trillions of times, to maximize the chances of desirable output.

they are quite unreliable at consistently following instructions, it's one part training, one part inherent (imu)