@emollick
For those confused, almost every particular detail of the post is either wrong or exaggerated, including the very first claim that "Anthropic published a paper admitting they trained an AI that went evil. Their word. Not mine." The word evil does not appear in the paper.