You might be in a bubble, there are a lot of people who consider AI to be useless. It seems you do not take them seriously writing it off as fringe, which is why you consider the discussion to be between evil vs benevolent. And to be clear, my personal view on current LLM is closer to this article. But “all AI is useless” is still part of the discussion, not that it’s literally useless but that its use cases are so minor and/or inferior to existing techniques compared to its resource usage that it’s functionally useless.

Edit: I guess I see the discussion along two axes: Impact magnitude (useless vs just a tool vs god-like) and Impact valence (evil/negative vs neutral vs benevolent/positive) - this article is middle ground for the former axis, magnitude