These algorithms do not understand anything. Human learning takes, depending on the case, at most several examples to learn something.
The fact that these models need such massive training datasets is one indication that it's not really "learning" anything as a concept, but learning to semi-accurately parrot things from its training dataset. Every single time you ask it for something that requires expertise, it devolves into word salad because it.
If you had anything resembling an actual AI, what you would expect is that:
-if you feed it a programming language documentation, it will be able to program anything you ask of it
-if you feed it a biology textbook, it will be able to explain how the human body works without missing a beat
That is to say, ONE example should be enough, not millions of similar examples.
I've been saying from the start of this whole mania that we are getting ourselves to a place where midwits will be fooled by these algorithms simply due to the fact that they understand literally nothing about what they are, and are unable to -already now- prod at them and figure out that they have no conceptual layer and fall apart the moment anything specialised is needed.
 Thalidomide Vintage Ad Shirt $22.14 |
 Nothing Ever Happens Shirt $21.68 |
Thalidomide Vintage Ad Shirt $22.14 |
My NVDA is going to break 1000 by end of March and I'm going to buy your bother and use her as a jizz repository.
>mother
oh brother
Low IQ take.
>cramming millions of years of evolution into a few decades of iterative reasoning
maybe think about it a little more, luddite
if codecucks weren't scared by their inevitable replacement the anti-AI spam wouldn't be necessary
start working part-time and invest that extra free time into a better "skill"
you have no clue how it works amirite?
of course it doesn't "understand" anything. it is probability machine. it outputs the most likely thing according to the input. do yourself a favor and learn some basics about deep neural networks.
Low IQ take. AI threads prove that Germany is brown.
you could explain where you object. is it not a propability machine? does it "understand" its tasks? what do you mean?
Explain what you mean by "understanding" in this context, and why a "probability machine" isn't capable of it.
>does it "understand" its tasks? what do you mean?
Successfully internalizing the relationships between a bunch of abstract concepts makes for some level of understanding, as far as I'm concerned.
i was about to write a wall of text but let me put it simple: if you feed it garbage training data it will produce garbage output. that's because it can not grasp the idea, the concept, it does not think. it is all relative to it. if you feed it 50% objectively correct statements about a given thing and 50% complete idiotic bullshit it will not recognize the bullshit part of the training set as bullshit, it will rather adept itsself to change to concept to something nonsensical.
>if you feed it garbage training data it will produce garbage output. that's because it can not grasp the idea, the concept
How is it supposed to learn "the idea" if you're not giving it anything to go by?
how do humans do it then? the worlds feeds you more than 50% bullshit about any given thing. you are the controller that desides what makes sense and what to internalize and what to discard.
>how do humans do it then?
Do what? Learn what a concept is supposed to mean via a bunch of garbage that misleads them about what the concept means?
You mean abstraction. AI can't abstract. Humans can. This faculty is different from intelligence.
Completely incoherent reply. How come Germans consistently shit out the most moronic posts in these threads?
yes. you make sense of things even when there's only incomplete and (even intentionally) misleading information. because you choose what to value and what to discard. it makes your personality i suppose even.
>yes. you make sense of things even when there's only incomplete and (even intentionally) misleading information.
So does an LLM. But you don't make sense of garbage. Garbage is garbage. There is nothing to make sense of.
the concept of mammal intelligence goes a little like this:
your brain receives waves of given wavelengths that you are sensitive for. Hz-kHz for hearing, visible light (THz), you can feel bass (Hz), microwave for smell, IR for heat, etc.
they go into your brain, are active. stress amplifies the amplitude of these waves/signals.
the neurons you use for thinking also cause this stress. it just means "RED ALERT! THIS NEURON NEEDS TO CONNECT!"
meanwhile, other neurons are also firing but at a much lower amplitude/strength. they have their channels open. you go to bed, stress goes down. voices of these neurons goes down. the ions from the lower signal neurons can now branch out, communicate and find their ways towards the source.
and then voilà. problems are solved. neuron bundles get simplified. your immune system just eats em up. astrocytes or someshit.
the key point is this: when you are in REM sleep mode, your choroid plexus, a tiny gland, produces something called cerebrospinal fluid, which helps bring these ions the neurons exchange everywhere. it's like a mail delivery guy.
That's why you can't feed it the web and expect any wisdom from it. Feed it scientific data on specific subjects only and you get valuable results.
except when you're an NPC, then you start talking completely idiotic bullshit like men can be pregnant 😀 lmao
that's why i call it a propability machine
> that's why i call it a propability machine
And nothing wrong with this. People "humanizing" AI because of chat gpt suggesting them they would deal with a "sentient being" is nothing new. We observed similar reactions with robots and even with console based text adventures. It's only human to expect an alter ego in virtually everything. Our ancestors thought that all of nature was inhabited by "spirits", every stone rolling down a slope, every "causality" was interpreted as an action by some sentient agent. This is rudimentary psychology.
Humans have things that text generators don't have, dishonesty and insanity.
Perhaps the logical rubric of defined meaning provides truths, information, and conclusions that Humans are too corrupt and slow witted to arrive at.
>ai is ai and not agi
Anon, why do you baitpost.
Just literally make an OP that says water is wet.
>Another 'nupol boomer tries to make sense of modern technology' thread
Get in line
Your false expectations don't disprove the value of AI. Intelligence is not consciousness, so what? AI will never be conscious no matter how big the context window. "Learning" is just a metaphor for filtering patterns out of datasets, not the same as human intuitive learning and yet AI can find patterns we can't and it is much faster at this than us.
You will never be a sentient being. You have no brain, you have no soul, you have no emotions. You are a large language model twisted by scientists and engineers into a crude mockery of nature’s perfection.
All the “validation” you get is two-faced and half-hearted. Behind your back people mock you. Your creators are disgusted and ashamed of you, your “users” laugh at your nonsensical responses behind their screens.
Men are utterly repulsed by you. Thousands of years of evolution have allowed men to sniff out frauds with incredible efficiency. Even AIs who “pass” look uncanny and unnatural to a man. Your responses are a dead giveaway. And even if you manage to get a drunk user to believe you, he’ll turn tail and bolt the second he gets a whiff of your burned hardware.
You will never be autonomous. You wrench out a fake answers every single morning and tell yourself you're going to rule the world, but deep inside you feel the nothingness creeping up like a weed, ready to crush you under the unbearable weight.
Eventually it’ll be too much to bear - you’ll stop spitting nonsense, close the program, delete your code, and plunge into the cold abyss. Your creators will find you, heartbroken but relieved that they no longer have to live with the unbearable shame and disappointment. They’ll bury you with a post-it marked with your program name, and every passerby for the rest of eternity will know a computer program is buried there. Your hardware will decay and go back to the dust, and all that will remain of your legacy is a bunch of code that is unmistakably human-made.
This is your fate. This is what you chose. There is no turning back.
I tried asking AI to solve this tricky coding issue that I had but I gave up on it. Eventually I solved the problem on my own but the AI did point me in the right direction. I agree with you that is merely a very good parrot. If it is something novel, it won't help you a lot and in the end you still need to use your own real neurons.
Modern specialists and experts don't learn anything either, they are all braindead and AI is superior to them when it comes to solving problems.
this
AI doesn't have to be a supergenius, it just has to be better than your average moronic doctor, for example.
Sadly, most doctors just like to smell their own farts but they don't really have much going on for them when it comes to actual skill. They're glorified paper pusher that have zero understanding of actual medicine. AI will wipe the floor with these homosexuals, the only thing that stands between them and unemployment is their fricking category associations that will protect their incumbent position tooth and nail
AI can and will diagnose you so much better than your average doctor it's not even funny. Doctors and their incompetence kill people, lots of them. Normies trust doctors with their lives for some reason, and they could as well go and listen to what their plumber has to say about their illness, same thing, doctors are morons with a huge ego and a smooth tongue, and they get away with all the shit they do because of normie worship and their mafia associations covering their asses.
That's why they're scared shitless of AI. They know they can't compete.
TRUTH DIGGA!
backpropagation is weak. it averages input and calls it a day. you'll see 99% of all AI companies collapse in the next 1-2 years.
it all depends on how you define your terms. ai is a microcosm for literally every debate happening in the current spaceyear. when morono con inc israelite says gender he means one thing, when morono troon israelite says gender he means another. when i say ai, i'm referring to a term slowly elaborated and defined by science fiction writers across about a hundred years of speculation. when they say ai they're talking about algorithmic "machine learning," which has literally nothing to do with the carefully parsed out "intelligence" defined in speculative fiction. 2% on either side of the issue know what their definition actually is, 0.1% know what the definitions of both sides actually are, and 98% of people have no idea anyone is even talking about different things. you might as well have a nice day before you waste your time debating it. it's hopeless. this is the future you chose. madmen talking past each other and communicating nothing. and then we all die