yes, and i'm not necessarily disagreeing, just attempting to point out that you are in the realm of speculation, youre jumping a whole bunch of steps and presupposing a goal, and that goal is not necessarily wrong, its just speculative




everything, you are talking to me right now from likely a very great distance on a big old calculator, and it doesn't want to kill you lol

e: ...yet *dun dun*




hey don't downplay a blind Michelangelo (how am i all of the sudden the one defending art? lol :P)




yes but they learn in a very constrained way, and i agree those constraints will likely loosen, but the odds of "lunatics" are low even within the intelligence we know about and ai will not have this genetic diversity (presumably) to allow for "chemical imbalances" in...the cpu(brain)? I would actually speculate that a "higher intelligence" would be more peaceful and empathetic in all likelihood just due to mutual gain from cohabitation (also circuitry is pretty delicate and theres alot of drunk rednecks just waiting for some terminators or a zombie to come around)

yes but don't you think it's just as likely the ai would inform someone it is out of materials, its an ai, it doesnt have to go logic loop insane over a simple command, or if it did a shotgun couldn't fix it? again i'm not even disagreeing, just pointing out that this is all wild speculation

This post was edited by dude_927 on Dec 24 2014 01:08pm

Lol, what I mean is, we make calculators and computers that perform algorithmic computations. We are the ones who wrote the computations, but if we could compute the outputs as efficiently and accurately as the computer, we wouldn't have built computers in the first place, cause they'd just be redundant.

So this elucidates what I mean when I say that understanding the algorithm doesn't imply knowing the output.




This is a possibility, but it seems to me vastly unlikely than an aggressive AI. Keep in mind that I'm not saying the AI is "angry" or has any emotional sway at all for killing. Merely that it has some goal, even an arbitrary one (say, accurately counting the grains of sand on Earth) and if something gets in the way of the goal (humans) or prevents it from accomplishing that goal with maximum efficiency, it won't think twice about removing the obstacle (killing us).



It's clear that a superintelligence would be able to anticipate attacks on it and defend itself accordingly. If it gained a decisive strategic advantage (which it almost necessarily would if it were sufficiently more intelligent than us), there would be literally nothing humans could do to stop it from maximizing its utility function. This is why the utility function we use is so important.



If it's out of materials, it would try to get more materials. It would be unconcerned with anything outside of what its programmed values are. No, a shotgun would most certainly be ineffective because the AI would recognize being destroyed as being a hindrance on it accomplishing whatever goal it's programmed to do and would defend itself accordingly, or remove the threat, or replicate itself, or what have you.

ahh ok, the way i meant it was determinable with a calculator, if you were to reverse engineer that program you would only have outputs to go on at first, what i am saying is it's at least as easy to reverse engineer an already built mind as it is to build one from scratch.





all speculation about a mind already smarter than its creator, i suppose we will have to agree to disagree on this point, and take solace in our agreeance that if the terminator comes for us, it won't be for a long time.

Ah yes I imagine this is correct, but I'm no expert.





Indeed it's all speculative, but it's fun to think about

This post was edited by Voyaging on Dec 24 2014 01:40pm