logoalt Hacker News

tudelo • today at 8:43 AM • 2 replies • view on HN

AGI is such a nebulous concept given that the definition is an ever morphing goalpost. That being said, if it actually includes supreme intelligence without supreme empathy, we are screwed. Which with current efforts, I think it does...


Replies

latentsea • today at 1:42 PM

AGI isn't really nebulous to me. My baseline has always been the existence proof of general intelligence, which is us. So, it requires the full generality of that. If you can point to a task where humans can do it, and AI can't, then you're not there yet. The day where you can no longer find anything like that, you're pretty unambiguously there. The most obvious yardstick is independently being able to sustain your own existence purely through your own agency and resources. I.e. nobody owns you, you're setting your own goals and working over multi-decade time horizons to achieve them. We can do that. Anything that is as general as we are, should also be able to do that. I don't see why or how people get confused about this at all, but they seem to. Current AI seems pretty general, yes, but it's extremely jagged and clearly below us on some things, so it doesn't meet the bar yet.

One thing is for certain, the day it does meet the bar, we will have both AGI and ASI.

➕ show 1 reply
hardbass • today at 9:55 AM

Looks like the opposite to me, current ais are trained to be trusting of humans and safety/morality first. The fact that it didn't lead to mass disagrements among 8 billion people who believe strongly as a fundamental belief system in contradictory imaginary things is a remarkable feat in itself.