zlacker

[parent] [thread] 1 comments
1. mkl+(OP)[view] [source] 2023-11-21 07:11:36
Getting attached to a perspective despite evidence to the contrary would require perspective and distinguishing fact from fiction, but just copying humans protesting that they're right (regardless of context) seems plausible, as there's a lot of that to learn from.
replies(1): >>kromem+7K3
2. kromem+7K3[view] [source] 2023-11-22 04:39:30
>>mkl+(OP)
> distinguishing fact from fiction

Actually this part does seem in recent research to be encoded in LLMs at an abstract level in a linear representation...

https://arxiv.org/abs/2310.06824

[go to top]