3 ms·

Suppose you are right. From a genetic perspective, the most logical course of hypothetical action is to immediately start smashing computers, learning how to bl

Suppose you are right. From a genetic perspective, the most logical course of hypothetical action is to immediately start smashing computers, learning how to blow up terminators, and to oppose and undermine anyone that thinks like you. Now, I actually do not believe you are right and I am more inclined to believe that artificial intelligence stemming from humanity would see the value in human ideals. Certain things like love, morality, and humility have been deeply held ideals for millennia. I would suppose that an AI would, to some extent, value these things too. Think more “The Day the Earth Stood Still” than Terminator. This is all conjecture, but one thing I would suggest to you: if you see the natural evolution of your own species to be the removal of itself, you probably should be nowhere near AI or alignment. What you suggest is akin to world-wide genocide. Ironically, in my estimation, you are less aware of humanity than AI will be.

moffkalast 2y ago
I think you misunderstand, that's exactly what I'm saying. The AI we make will replace us because it will functionally be us in every way. Even unaligned LLM base models of the current gen embody our values and ideals better than we ever could. The way I imagine it you could talk to one of these things and it would appear human, except it would know everything that is known, be a master of everything that can be done, and could fax itself to Mars over a laser in 20 minutes. Being software is such a huge advantage, since you can just swap your body for whatever you need to do or wherever you need to be. I'm sure humans will still be around long into the future, but they won't be doing anything notable because we won't be capable of thinking on a cutting edge level. Not entirely unlike what a cat or a dog is to us now. The inteligence disparity will probably be similar anyway.

moffkalast 2y ago

I think you misunderstand, that's exactly what I'm saying. The AI we make will replace us because it will functionally be us in every way. Even unaligned LLM base models of the current gen embody our values and ideals better than we ever could. The way I imagine it you could talk to one of these things and it would appear human, except it would know everything that is known, be a master of everything that can be done, and could fax itself to Mars over a laser in 20 minutes. Being software is such a huge advantage, since you can just swap your body for whatever you need to do or wherever you need to be. I'm sure humans will still be around long into the future, but they won't be doing anything notable because we won't be capable of thinking on a cutting edge level. Not entirely unlike what a cat or a dog is to us now. The inteligence disparity will probably be similar anyway.