Comment on It's not just statistics: GPT-4 does reasonparentComments−flandish3yUsing the metric “can reason” on a LLM is like using the metric “can bleed” on a stone.Maybe some red stuff comes out when you break it. Is it blood, or is it something pumped into the other side?
Comments
Using the metric “can reason” on a LLM is like using the metric “can bleed” on a stone.
Maybe some red stuff comes out when you break it. Is it blood, or is it something pumped into the other side?