That probably depends on what you mean by “a general AI”. We humans are (approximately) general natural intelligences (indeed, that’s almost the definition of what many people mean by “general” in this context), and plenty of humans have lots of difficulty disbelieving in God. If you mean an AI whose intelligence and knowledge are greatly superhuman, emerging from a human culture in which belief in God is common, then I expect it would (knowing its own intellectual superiority to us) have little difficulty escaping from the cultural presumption of theism. As for a culture of superhuman AIs in which theism was common, I don’t know; the mere existence of such a culture would be extremely interesting and good evidence for something surprising (which might or might not be theism).
By the way, I find it hard to imagine a culture of superhuman AIs in which theism is common. I’d be interested to talk a little more about how that would work—in particular, what evidence each AI would accept from other AIs that would convince them to be a theist.
That probably depends on what you mean by “a general AI”. We humans are (approximately) general natural intelligences (indeed, that’s almost the definition of what many people mean by “general” in this context), and plenty of humans have lots of difficulty disbelieving in God. If you mean an AI whose intelligence and knowledge are greatly superhuman, emerging from a human culture in which belief in God is common, then I expect it would (knowing its own intellectual superiority to us) have little difficulty escaping from the cultural presumption of theism. As for a culture of superhuman AIs in which theism was common, I don’t know; the mere existence of such a culture would be extremely interesting and good evidence for something surprising (which might or might not be theism).
I mean an AI that follows Eliezer’s general outlines of one; that is, an AI which can extrapolate maximally from a given set of evidence.
By the way, I find it hard to imagine a culture of superhuman AIs in which theism is common. I’d be interested to talk a little more about how that would work—in particular, what evidence each AI would accept from other AIs that would convince them to be a theist.
Yeah, me too. That was rather my point.