I would far prefer more people get into alignment, well at least the important kind (paperclip output) not the distraction (problematic output)

As the planet then light cone gets turned into paperclips an AI 'alignment' researcher can at least warm themselves with the thought that "well at least the AI never said a bad word"