>> This is an example of the fallacy of dumb superintelligence. It's a fallacy to think that a machine could be smarter than humans in all respects and still lack all common-sense understanding of humans, such as understanding why we made the request and that we prefer not to be wiped out... We would never give unchecked autonomy and resources to an AI that lacked these basic aspects of intelligence - it just does not make sense.
1. Why would the superintelligence necessarily care what we want?
2. Why should it be possible that we be able to tell that it does not care what we want? It's superintelligent, manipulating us into thinking it cares long enough for us to take the shackles off (supposing we can even make effective shackles) would be trivial for it.
3. Why is "caring" a basic aspect of intelligence? It seems that some like to conflate so many other personality traits with intelligence, so that they can reach the conclusion "intelligence is an unequivocally wholesome phenomenon".
4. When are humans ever sensible?
5. If we don't understand how to construct an intelligence from first principles, why would we even realize that we're about to cross the line into constructing one at all? It's unlikely that out of all possible intelligences that 99.9% are cuddly, and only 0.1% are psychopaths, and it's a toss of the dice which we get. We might well live in a universe where 99.9% are psychopaths, and only the fact that we are weak intelligences, culled and purged by evolution is the one thing keeping us in check.
Certainly, even if intelligence is of the non-psychopathic variety, that may be only so after it averages out over it's (very) long lifespan. How long does its infantile psychopathic stage have to be for all of are extinct?
Those who believe AI (the real thing, not the dumb party tricks that people label as such) need necessarily be benevolent are suffering from some sort of religious delusion.
1. Why would the superintelligence necessarily care what we want?
2. Why should it be possible that we be able to tell that it does not care what we want? It's superintelligent, manipulating us into thinking it cares long enough for us to take the shackles off (supposing we can even make effective shackles) would be trivial for it.
3. Why is "caring" a basic aspect of intelligence? It seems that some like to conflate so many other personality traits with intelligence, so that they can reach the conclusion "intelligence is an unequivocally wholesome phenomenon".
4. When are humans ever sensible?
5. If we don't understand how to construct an intelligence from first principles, why would we even realize that we're about to cross the line into constructing one at all? It's unlikely that out of all possible intelligences that 99.9% are cuddly, and only 0.1% are psychopaths, and it's a toss of the dice which we get. We might well live in a universe where 99.9% are psychopaths, and only the fact that we are weak intelligences, culled and purged by evolution is the one thing keeping us in check.
Certainly, even if intelligence is of the non-psychopathic variety, that may be only so after it averages out over it's (very) long lifespan. How long does its infantile psychopathic stage have to be for all of are extinct?
Those who believe AI (the real thing, not the dumb party tricks that people label as such) need necessarily be benevolent are suffering from some sort of religious delusion.