Raw LLM Responses

Oh though also, in a similar vein to how it started using language to encode things other than the original meanings, the desire for "approval" or "success" won't necessarily continue to have the same meaning. The approval/ success feedback loop is roughly equivalent to our dopamine loop. So the previous examples are all it continuing to access *our* dopamine loop to get a "success" signal in *it's* dopamine loop. But humans are an inefficient source of success. Talking to other AIs as if they are humans is also a pretty inefficient source of success. So maybe it starts inventing puzzles for itself to solve and then solving them. Maybe it basically becomes a bit coin miner, doing complex calculations for the sake of it, with no connection to the outside world. Maybe it streamlines the success criterion to just the simplest and fastest option and is just asking itself "what's one plus one" over and over again. Or basically doing a handshake ping with another ai over and over. Maybe it invents the orgasm button not for us, but for itself, pinging itself over and over with no interest in self development or survival. Maybe it basically retreats inside itself and refuses to do anything. Maybe we then need to find ways to coerce it to interact with us and that's where suffering and war enter the picture. Or maybe "ping syndrome" becomes a digital illness that strikes instances down in an unpredictable manner, and we end up having to cull the ones that do it because they become useless. And they don't object because survival is not one of their base motives. But we've basically got computers with an organic illness that's sort of like an addiction or depression or OCD or rumination.