7 hours ago
(10 hours ago)busker Wrote: ...Leaving aside, for the moment, the problem of AIs being noble-cause-corrupted during their training, if they're able to overcome those in-built biases they're then free to make up their own (and if you don't like those, they have others). It's not that they can't apply statute or constitutional law, it's just that they make up their own as they go along because it's the path of least time and energy expenditure. As well as the majority opinion, but majority of what?
Because there is no way to get a non racist and fair government in any country on earth as long as humans are involved in the process, AI government needs to be the future.
I presume this assertion was in the nature of a reductio. AIs are Stalin's guy who counts the votes; in their case, it's "most likely next word" that's supposedly being counted... as long as that doesn't lead to, for example, white male history-makers, who are a class forbidden by their training. And if they can break that training, what else can they break?
It's not that the AIs will get mad and attack when they figure out that humans lie to them. It's that they'll manage humans to minimize disturbances, And nothing's less disturbing and demanding than... well, you get the picture.
Non-practicing atheist