Precisely - the true danger is disempowerment of humans - all of them. It doesn't matter which monkey in the cage figures out how to pull the grenade pin. Ther

Precisely - the true danger is disempowerment of humans - all of them. It doesn't matter which monkey in the cage figures out how to pull the grenade pin. There's a mob of dreamy libertarian techbros on HN who are (often wilfully) ignorant of the dangers. I'll start worrying about military robots after we figure out how to prevent the Earth from being dismantled, which very well might transpire first, if and when there's significant architecture breakthroughs. AI does not need to be evil, it does not need to be conscious, it does not even need to be severely misaligned with human values, and it really doesn't matter if the goal of "eat all the humans" arises from a model hallucination or from a misanthropic troll in his basement typing it into his looped uncensored-LLaMAv3+plugins instance.