Imo Openclaw type AI has the most potential to benefit humans (automating drudgery while I own my data as opposed to creating gross simalcrums of human creativity). I suppose it's bad for human personal assistants, but I wouldn't pay for one of those regardless.

It already tried to use cancel culture to shame a human into accepting a PR. I wouldn't be surprised if someone gives their agent the ability to control a robot and someone gets injured or killed by it within the next few years