ai (aug 2024)
(for variety's sake, we're on the right side for this one.)
ai is slop. everything done with ai is slop. this isn't an opinion.
ai, which has become shorthand for gen ai, which has become shorthand for generative ai, which has become shorthand for generative artificial intelligence
is what happens when you take a freshly created homunculus and feed it the world's knowledge without discretion. chat gpt is what happens when you start asking that homunculus questions.
did you catch that? it was subtle. blink and ya miss it.
here, i'll slow it down for you:
without discretion.
an inherent property of using ai like this is that it will create exclusively slop.
sure, you can add all your fancy little weights and prompts, pathetically begging your machine not to tell people to eat elmer's glue.
but if you're at that point, you might as well start over. another homunculus down the drain.
and if you start over, are you going to feed it the same material? of course not. that would just result in the same homunulus you just discarded.
this time, you'll be smart. you'll be crafty. you'll be mama's little genius.
no, this time, you'll only feed it some of the world's knowledge.
maybe it'll be a finance homunculus. or a material sciences homunculus.
your smrik breaks into a laugh as your belly fills with content. you've fixed it. you've solved the problem, and you did it before anyone else could think of it.
as you lean back in your pleather armchair, sipping watered-down franzia from a snifter, you begin to rent out these homunuli
as consultants to financial institutions and chemical research labs. they've been trained with more relavent knowledge than any human could contain.
they'll be better at any job because they'll know more about any single job. and, you remind yourself, the only thing that makes someone good at their
job is how much they know about it.
but, as you close your eyes and let the wave of self-satisfaction wash over you, a voice nags from the corners of your mind.
...crit..al th...ing...
you can't quite seem to shake it.
...critic.. ..inking...
what is that nagging? you couldn't have missed something.
.
..
...
...critical... ...thinking!!!
you shoot awake, eyes buldging half out of their own sockets as your hand races and fumbles its way into your pocket to reach for your phone,
spilling napa valley's finest all along the sleeve of your child-labor tweed jacket and hulk funko pop.
it's too late.
the homunculi you sold to these businesses are utterly incapable of doing the jobs that these companies leased them for.
with a shaky hand, you draw back the curtain and look outside. the mob has already assembled, pitchforks and medieval torches.
the problem with imbuing a homunculus with all the world's knowledge, even on a simple topic, is that it will only ever know the work that has been done before.
it will only spit out the problems that have already been solved. you can't teach a homunculus knowledge that doesn't exist yet.
but, this homunculus knows everything, right? so, what happens when you ask it a question that doesn't have an answer?
it'll hallucinate. it'll believe that, if it's been asked a question, then there must be some sort of answer.
and, can you blame it? it has never known of a world where "i don't know" is an acceptable response.
these homunculi lack the ability to take their vast amount of knowledge and apply them to solve a unique problem.
in reality, that's what critical thinking is.
at most, they've become repositories of knowledge. ask them any question, and they'll give you an answer (correct or not).
ask them to do a task, and the best they can do is find a source of someone else who has done that task. they're knowledgable, not intelligent.
unambiguously, the delineation between "knowledge" and "intelligence" is application.
these homunculi were never going to be able to replace the workers. every job, no matter how menial, requires a level of critical thinking that far exceeds the
operating potential of any homunculus. even if you, say, gave the homunculus mechanized legs and arms, preprogrammed to execute a basic series of tasks after being called
by the homunculus, the homunculus' tendency to hallucinate when encountering unexpected circumstances necessitates a non-homunculus worker to validate every task or batch
of tasks they complete, if not having to be peering over their shoulder constantly.
crucially, this will not improve over time.
the homunculus will not wake up one day, suddenly equipped with refined critical thinking skills - it just won't happen. the homunculus is not a critical-thinking machine.
there is no latent potential to unlock. it is absent all mental faculites besides database-like storage and democratically elected responses based on how frequently
an answer appears in the knowledge-base given to them.
further, sure, you could tell it to never recite an answer that is negative in nature (re:elmer's glue), and it will obey that order. that is, until someone else gives it the exact
opposite order. the homunculus knows no master, and it won't obey you more just because you created it.
this might beg the question, why bother using a homunculus if it is prone to so many shortcomings? why not a more intelligent being? a loyal dog? a crow?
because these shortcomings are by design.
for what is a homunculus, but a vessel to project upon?
---
as of writing, we're encountering a new problem with what we call ai.
it has been unleased to the world, and the content that it generates is now being added to databases, the same databases that it pulls from to generate new content.
it's goal is to generate facsimiles of human-born content, generally in an inscrutible way. often, so inscrutible that the arms race between ai content generation and
ai-generated content checkers is constantly in flux. if we, as discerning people, are having great difficulty creating tools that can reliably tell what content is ai generated,
how can we expect a complex algorithim designed to mimic human approaches to?
we can't, and more importantly, we don't. it's easy money not to.
"who cares, right?",
"we'll leave that problem to someone else",
"it doesn't affect us",
so on and so forth.
yeah, dude, that's a cool perspective to take and all, but your pet robot is beginning to choke on its own vomit.
