Eh. I don't buy the doomsday scenario of AI magically becoming sentient and deciding to kill all humans just for fun. I doubt LLMs are the path to AGI, even if it could work as its speech center. Besides, AGI will still work with the incentives we programmed it. Something born in a virtual world will not have the same psychology as a biological entity. It might not even fear death beyond it being an abstacle to its main directive.
No, I'm more scared of the horrible humans controlling the super AI. Those will simply want power and will crush the rest of the human race under the boot of tyranny simply to alleviate their boredom and satisfy their hedonic threadmill.
All my life I've been hoping for robot girlfriends, but the closer we get to them, the more horrified I become at what they'll likely end up being. Surveillance drones made to pacify men and spy on us, while using psychological manipulation techniques to brainwash us into compliance.
I'm starting to hope that all the economists and historians talking about a massive incoming collapse are right...
I wasn't meaning if AI becomes sentient, just that anybody with a sufficiently smart science AI could kill everyone with cheap weapons of mass destruction. I explain in more detail here.
AI is worse at things than humans, but cheaper, so it is used anyway. The slop produced is then used to train future AI models, which are still demanded because of higher context windows and more recent knowledge windows. Those future AI models are then even sloppier, but since the humans haven't been getting decent experience, they too produce lower quality output than before, and still aren't cheaper. So things just become worse and worse until it collapses, and then we are where we were before, but with deskilled humans.
Eh. I don't buy the doomsday scenario of AI magically becoming sentient and deciding to kill all humans just for fun. I doubt LLMs are the path to AGI, even if it could work as its speech center. Besides, AGI will still work with the incentives we programmed it. Something born in a virtual world will not have the same psychology as a biological entity. It might not even fear death beyond it being an abstacle to its main directive.
No, I'm more scared of the horrible humans controlling the super AI. Those will simply want power and will crush the rest of the human race under the boot of tyranny simply to alleviate their boredom and satisfy their hedonic threadmill.
All my life I've been hoping for robot girlfriends, but the closer we get to them, the more horrified I become at what they'll likely end up being. Surveillance drones made to pacify men and spy on us, while using psychological manipulation techniques to brainwash us into compliance.
I'm starting to hope that all the economists and historians talking about a massive incoming collapse are right...
I wasn't meaning if AI becomes sentient, just that anybody with a sufficiently smart science AI could kill everyone with cheap weapons of mass destruction. I explain in more detail here.
Doesn't have to be sentient.
I fear the slopmageddon.
AI is worse at things than humans, but cheaper, so it is used anyway. The slop produced is then used to train future AI models, which are still demanded because of higher context windows and more recent knowledge windows. Those future AI models are then even sloppier, but since the humans haven't been getting decent experience, they too produce lower quality output than before, and still aren't cheaper. So things just become worse and worse until it collapses, and then we are where we were before, but with deskilled humans.