To add a little nuance and a bit of a detour from the original topic, some Rationalists (I'm thinking Scott Alexander) tend to spend a lot of brainpower on negative aspects of AI too - think the alignment problem.

The category of events having near infinite positive or negative outcomes with zero to few examples where it's difficult to establish a base-rate[prior] appears to attract them the most. Conversely, an imagined demonic relationship with a yet to be realized unaligned AI results in a particular existential paranoia that permeates other enclaves of Rationalist discourse.