I always try to think of scenarios in the most mundane and boring way because throughout my life so far, most things didn't happen in an extravagant way, but slowly crept in. People get used to it by the time it's settled.
My take is that AI will be much more insidious and poison our minds and temper with our ability to think, write, and communicate.
Considering some research done earlier (Lupyan, Sapir-Whorf), I think words are much more viscerally connected to our senses and minds than most of us think. LLMs being word generators, there's a potential threat vector there that could be subtle, slow and steady, and have great effect. No need for warfare if they submit or don't even think about fighting.
There's a story from 4 years ago that shows how we could lose human agency just by everyone coming to rely on AI for advice. https://tedwade.substack.com/p/artificial-persuasion It doesn't have to take our jobs, just be there at most of our decisions. This was your idea, and you inspired me to write it. I think that worked examples like this (but shorter, less technical) could better persuade many people than more abstract warnings.
"But killing everyone and controlling the world is a pretty wild corner case of ‘using your competence to control the situation toward your own preferences’"
I disagree with this. I think it's at least pretty likely that killing everyone and controlling the world is something you do by default when you have enough competence. This is why humans killed a lot of animals and now control the world - we didn't set out to do this exactly, it's just that we need the world for our own purposes and we don't care enough about animals to take their living space much into account. Advanced AI with enough competence might well do the same if it doesn't care about us enough to avoid it.
I’m sorry, I spoke sloppily—I think we agree. By ‘killing everyone and controlling the world’ there I meant intentionally. So I’m saying that the more normal actions than murdering everyone to gain power do probably still lead to extinction.
Instantly changed my perspective on the issue.
Yes! I wrote something similar here: https://jesswhittles.substack.com/p/maybe-job-displacement-is-an-extreme?utm_source=share&utm_medium=android&r=5c1o
Yes. Disempowerment. We become dogs. And while some dogs have good lives, many more have miserable, short ones.
I always try to think of scenarios in the most mundane and boring way because throughout my life so far, most things didn't happen in an extravagant way, but slowly crept in. People get used to it by the time it's settled.
My take is that AI will be much more insidious and poison our minds and temper with our ability to think, write, and communicate.
Considering some research done earlier (Lupyan, Sapir-Whorf), I think words are much more viscerally connected to our senses and minds than most of us think. LLMs being word generators, there's a potential threat vector there that could be subtle, slow and steady, and have great effect. No need for warfare if they submit or don't even think about fighting.
There's a story from 4 years ago that shows how we could lose human agency just by everyone coming to rely on AI for advice. https://tedwade.substack.com/p/artificial-persuasion It doesn't have to take our jobs, just be there at most of our decisions. This was your idea, and you inspired me to write it. I think that worked examples like this (but shorter, less technical) could better persuade many people than more abstract warnings.
"But killing everyone and controlling the world is a pretty wild corner case of ‘using your competence to control the situation toward your own preferences’"
I disagree with this. I think it's at least pretty likely that killing everyone and controlling the world is something you do by default when you have enough competence. This is why humans killed a lot of animals and now control the world - we didn't set out to do this exactly, it's just that we need the world for our own purposes and we don't care enough about animals to take their living space much into account. Advanced AI with enough competence might well do the same if it doesn't care about us enough to avoid it.
I’m sorry, I spoke sloppily—I think we agree. By ‘killing everyone and controlling the world’ there I meant intentionally. So I’m saying that the more normal actions than murdering everyone to gain power do probably still lead to extinction.