I'm referring to actual people I argued with in the past. People convinced that AI in a self-driving car would involve the car calculating whether to kill a pedestrian or the driver, rather than trying to figure out whether this thing half obscured by foliage is a speed limit sign or not.

Obviously that's not what everyone argues, my point is that there's a lot of chaff in such arguments and not much wheat. People make a lot of noise about dramatic but completely unrealistic scenarios, while ignoring the far more boring reality.

The PauseAI people are for instance talking about human extinction, somehow. And not crappy GitHub PRs.