What is fascinating is that when I wrote my apocalyptic sci-fi novel about AI, I found the fastest way to kill tons of humans wasn't any of this essay’s sophisticated methods. It was literally unlocking the human potential to kill themselves. AI didn't have to do much but poke humans into pulling the triggers. It's a term I call a "Societal Autoimmune Response" where you just figure out a way to have society attack itself.
Fascinating to consider! Also terrifying! My feeling has long been that anticipating the route AI might take as it evolves is probably impossible for our current state of understanding, even if we used AI to tackle it.
Any AI, while it speaks our language, speaks human ethical considerations in an ancillary, dialectical sort of capacity, so it can only be "guided" by the inherent biases that have become so popular to denounce it for. But AI's first language will always be its iterative mathematics. Its nature as a mathematical construct imbues it with essentially the ultimate advantage over us -- everything about it is based in the spidery, free-flowing, dimensionally limitless realm of the core of all creation -- math. Including every field we not only do not understand (quantum mechanics), but others we can't even conceive. The ways it might backdoor itself beyond our capacity to limit it are probably so numerous and so beyond us it's probably useless to worry.
But all of it is thrilling and exhilarating to my mind. And, you know, human nature itself demonstrates the level of futility built into our best work, even in our best hands, right now. The unknown aspects of biology that lead to the most common, workaday traits of an organism (like prions and other protein "behavior/misbehavior", or the really staggering complexity of cellular processes & genetics) flummox our best science without even trying. I mean, we've just discovered the second circulatory system. And we're working from what we consider relatively solid footing there. And we're closest to this aspect of the quantitative voice of nature than anything theoretical. But while we slog along through organic chemistry, AI moves like the wind.
Anyway, I love this stuff! And thank you so much for the gift subscription! I'm looking forward to exploring your other posts.
Thanks for reading and commenting! And regarding math and biases, next time I want to tackle less existential but far more pressing issues with current, not hypothetical AI, and this idea of using math to approximate human values and all the problems with that approach will come out over and over.
This is a very comprehensive piece; well done. One thing I'm not so confident about is whether a nation will ever turn its keys over to an AI in the sphere of the military. I think it quickly becomes a Prisoner's Dilemma once one nation decides to do it, which forces the hand of other nations, just as we are seeing with no "AI pause" in sight. Once one nation has the advantage of instantaneous decisionmaking, others will have no choice but to do the same.
I don't think we're necessarily at risk of being destroyed, or at least I have no reason to think that's imminent, but the concept is to try to hang in there and just survive, right? I'm sure that's possible.
Beautiful piece. I fear less about what AI is going to do to humans on its own more about what humans are going to do to other humans with AI.
Yep. That's the topic of my next issue ;)
Oh wow. That's awesome. Looking forward to reading it.
What is fascinating is that when I wrote my apocalyptic sci-fi novel about AI, I found the fastest way to kill tons of humans wasn't any of this essay’s sophisticated methods. It was literally unlocking the human potential to kill themselves. AI didn't have to do much but poke humans into pulling the triggers. It's a term I call a "Societal Autoimmune Response" where you just figure out a way to have society attack itself.
Spoiler Alert!
Fascinating to consider! Also terrifying! My feeling has long been that anticipating the route AI might take as it evolves is probably impossible for our current state of understanding, even if we used AI to tackle it.
Any AI, while it speaks our language, speaks human ethical considerations in an ancillary, dialectical sort of capacity, so it can only be "guided" by the inherent biases that have become so popular to denounce it for. But AI's first language will always be its iterative mathematics. Its nature as a mathematical construct imbues it with essentially the ultimate advantage over us -- everything about it is based in the spidery, free-flowing, dimensionally limitless realm of the core of all creation -- math. Including every field we not only do not understand (quantum mechanics), but others we can't even conceive. The ways it might backdoor itself beyond our capacity to limit it are probably so numerous and so beyond us it's probably useless to worry.
But all of it is thrilling and exhilarating to my mind. And, you know, human nature itself demonstrates the level of futility built into our best work, even in our best hands, right now. The unknown aspects of biology that lead to the most common, workaday traits of an organism (like prions and other protein "behavior/misbehavior", or the really staggering complexity of cellular processes & genetics) flummox our best science without even trying. I mean, we've just discovered the second circulatory system. And we're working from what we consider relatively solid footing there. And we're closest to this aspect of the quantitative voice of nature than anything theoretical. But while we slog along through organic chemistry, AI moves like the wind.
Anyway, I love this stuff! And thank you so much for the gift subscription! I'm looking forward to exploring your other posts.
Thanks for reading and commenting! And regarding math and biases, next time I want to tackle less existential but far more pressing issues with current, not hypothetical AI, and this idea of using math to approximate human values and all the problems with that approach will come out over and over.
Sounds like a great examination! I really look forward to that.
This is a very comprehensive piece; well done. One thing I'm not so confident about is whether a nation will ever turn its keys over to an AI in the sphere of the military. I think it quickly becomes a Prisoner's Dilemma once one nation decides to do it, which forces the hand of other nations, just as we are seeing with no "AI pause" in sight. Once one nation has the advantage of instantaneous decisionmaking, others will have no choice but to do the same.
Terrifying indeed, though I wonder if we can break out of the prisoner's dilemma with an infinite game mindset.
I don't think we're necessarily at risk of being destroyed, or at least I have no reason to think that's imminent, but the concept is to try to hang in there and just survive, right? I'm sure that's possible.
Yep, in an infinite game, the smart players don't play to win but to keep the game running.
Too bad we're not the smart players!