Narrow finetuning can produce broadly misaligned LLMs

discussed 2 days ago: https://news.ycombinator.com/item?id=43176553

Is anybody pointing on the fact that "alignment" is brain-washing?