Comment by root_axis

>it shouldn't be a stretch to believe that sentience might produce answers that people prefer

Even if that were true, there's no reason to believe that training LLMs to produce answers people prefer leads it towards sentience.