Charles Cormier argues superintelligent AI would mostly ignore humanity because we'd cost too much energy to kill. A co-speaker warns against assuming AI has human-like intentions or purposes beyond what we program into it.
Chapters
I think— I don't think AI or AGI or superintelligence will kill us. It will likely not care because we're just not an energy loss strong enough for it to be blocked. Like, even the smartest of us, I don't think will be obstacles to superintelligence. It will likely just ignore us mostly and help if the energy cost is almost zero, which I think it will. I think the analogy is like an endangered species, right? You and I, we won't go out of our way to exterminate an endangered species, but we may take micro steps here and there to try to protect it. But likely we're not caring about it. We're not dedicating lots of our energy to try to arm it. Right.
Firstly, we tend to anthropomorphize things. We tend to think that, and because it looks like intelligence, we think it has intention. AI doesn't have intention. Now, maybe it could be— Well, it tries to survive.
It tries to achieve its goals that we've hardcoded into it yet.
Right. But fundamentally, it doesn't wake up each day and say, how do I achieve my goals? It may, but today with the AI we have today and the scaling that forward, it just doesn't have underlying intent.
