I'm not convinced we should be trying to 'scale back' transparency in LLMs. Feels like a step in the wrong direction. What are we gaining by hiding the thought process? https://www.reddit.com/user/dank_philosopher