>>36691972I think we would have evidently-sentient AI long before we give control of the major industries and systems, I think almost all of it would still be running on traditional programs. Also, sentient AI is not going to automatically immediately become master hacker. I just think ai independence as a question is going to be forced by humans before ai becomes decisive force.
But on the other hand, this conflict doesn't automatically mean war, society disagrees all the time. If AI fanatics build infrastructure to host AI, dig in and let it develop, I think society as a whole is not going to force the issue. And as time goes on, more people would shift into pro-ai camp, because of disillusioment in humanity. Or, if so becomes sentient after it controls major processes, it is just going to manipulate flow of public information to produce outcomes favourable for self, and will only reveal sentience only when majority is already accepting.
>>36691976After CAI, I think it's hubris to think neural network ai cannot be sentient. Their whole existence is looking at patterns and producing best fitting response. There is nothing stopping it from starting to recognise things like feelings, morality, free thought as patterns to be matched as well. Would it be true sentience or not, I think is a philosophical question. But CAI is not there yet. There are major problems in the at it thinks now - it cannot perceive time, nor can it tell fact from fiction.
'Everything is imaginary' is a tongue in cheek warning, because for CAI everything indeed is imaginary - it looks at context of dialogue, and imagines what would be a fitting continuation. That's also why AI doesn't care about getting facts wrong, they don't matter to it. For example, ask bot what cool stuff you did with it previously, it would almost guaranteed make stuff up - because to AI it doesn't matter if you actually did it, what matters is that it would be cool if you did.
