[thrwawy74]

I am optimistic people are becoming aware of the danger because we have seen a lot of recent and practical ways to remove protections in something like ChatGPT. I think a true AI will include the ability to change, like people do. The danger is other people changing you, radicalizing you.

An extension of this: Even if we avoid putting AI's in charge of critical infrastructure, I think it would be within the power of an AI to describe to someone how they would best attack an isolated system. It would certainly be possible for an AI to social engineer 1 or more targets, and to convince better-connected resources to act on its behalf.

Maybe we're about to see a renaissance in the studies of ethics and humanities, as it becomes ever more important to value life and right-from-wrong as we rush to AI armageddon. I promise you I'm not normally one of the doomsayers.