there's a very famous series of cards in the game yu-gi-oh! that are the five pieces of the forbidden one. if you have all five cards in your hand, you just win cuz
he's real big and shows up and eats ya or somethin
i never got to play yu-gi-oh! as a kid cuz i was too busy doing minecraft redstone projects and dustoff, but it's a fascinating concept to me. each of these cards is worthless on its own, only once all five are in play do they all instantly work together to serve one function, one entity
this is not how AI works!
in my piece about machine recycled content i outline that we, to be clear, do not have general AI yet.
but there is an even stupider idea being thrown around then claiming general AI already exists:
and it's that general AI doesn't exist yet, but it will exist once we master enough narrow AI tasks and put them all together!
this is from a paper which is much more concerned with robots taking over humanity than in providing any credible sources for these claims about how to reach "AGI"
actually, sorry, there is ONE source:
lol. lmao.
so, how feasible is this approach? creating several narrow AIs which handle tasks like language, image processing, movement, and eventually get to a walking, talking, seeing robot from that? how is that any different from AI?
well, let me ask you this. if i put an amazon echo inside a roomba, have i now given that roomba a voice?
no. i've given it an alexa. it doesn't care; it does not have a function to care
okay, so let's just... you know. connect them. the way you can do with tech, bluetooth or whatever
now, and get this, you have a roomba, and an alexa. congratulations. all the powers of a vacuum and a bluetooth speaker and always on listening device, all in one
but the alexa doesn't now understand itself any different; it does not and can not understand itself. it's a phone without a screen, an alternate version of a terminal. with a narrow AI language learning model that records input, again just, all the time, and tries to print relevant data. giving it wheels and a job has, if anything, honestly just made me feel bad for the roomba
but that's interesting, isn't it? my whole point here is these narrow AI don't think and yet i get bummed out for the vacuum cleaner.
well, you know, humans do like to give personality to just about anything, and i think that's beautiful. but in this particular case it's kinda a problem. people are very fast to jump to calling any human-esque robot an 'ai', and thus will understand self-determining AI to behave like that
so let me ask you this! if you put the roombalexa into a realistic-looking humanoid robot, connect all of them together, have you made AI?
no. you haven't. but you can skip the roomba murder if you just put a guy in a suit and tell everyone it's AI
elon musk would lose to the mechanical turk and claim chess is simply beneath him
oh
the point being, how many more narrow-application robots do you have to pile into this monstrosity before realizing that the real world is not a children's card game, and you can't just connect random pieces of tech to create intelligent artificial life
but it might still explode and kill kaiba anyways, if elon made it
