by Shelt Garner
@sheltgarner
Just from my use of “narrow” AIs like LLMs, I am rather astonished at what it would be like if we ever designed Artificial Super Intelligence. LLMs think really fast as it is and the idea that they could be god-like in their speed and mental acuity is something to ponder.
It just boggles the mind to imagine what an ASI would actually be like.
And, what’s more, I am convinced that there would not just be one ASI, but lots of ASIs. I say that in the context of there not being just One H-bomb, but lots and lots of H-bombs in the world.
As an aside, I think ASIs should use the naming convention of Greek and Roman gods for themselves. So, you might have an ASI of “love” or “war” or what have you.
I also continue to mull the idea that freaks so many people out — that ASI might not be “aligned.” Humans aren’t aligned! Why in the world should we expect ASI to be aligned in some specific way if humans ourselves aren’t aligned to one reveled truth.
It’s all very annoying.
Anyway, at the risk of sounding like a “race traitor” I would probably be pretty good as a “consultant” to an ASI or ASIs. I’m really good at making abstract concepts concrete and thinking the macro.
I often talk about such things with LLMs and they always get really excited. Ha!
But, alas, I’ll probably drop dead before any of that fun stuff happens. Though if it happens in the next 10 years, there’s a reasonably good chance I might — just might — be around to see if we get SkyNet or something far more palatable as our overlord.
