[trade_monkey]

RLHF is a band aid on not having enough data that fits your own biases and answers you want the model to give.

[astrange]

It won't give answers at all if you don't train it to. It will output more questions because that's a more obvious completion to an incoming question.