so i was cooking this randomass theory on ai, and it reminded me of an anime by netflix PLUTO. for those who haven't watched it, it's okay to be uncool, i would still give you the synopsis. pluto is about a robot who's killing the 7 most powerful robots, that's it.
but what it introduces is so cool, it gives us an idea, what if we add personality in one ai, like all of them, so for that ai to choose a suitable personality it'll take a huge time and maybe probably never wake up, and if we introduce negative emotions then only this thing can be fixed.
now this is all my theory on AGI, what if we do the same with ai models? i mean i get it, emotions are very hard to define in actual ai research, but what if personality is all we're missing?
hear me out. i personally see only 2 good ways to make ai actually think novelly.
1st one is when we give ai a world, like how we humans got to experiment with chemicals, machines and what not. ai data is pretty much pre-experimented, we're not giving it a 3d simulation world to play around, and if we do that we have so much higher chances to get some truly cool AI.
2nd is yup, pluto, we need something similar with that Affective Computing, if we somehow add a variable of emotions we can see our ai would have direction instead of just taking instructions like a good boy.
either way we're basically trying to build a kid and skip the growing up part. maybe that's the whole problem.
i know my ideas might be sounding weird but that's okay, remember it's okay to be uncool.