what is the point in having an AI comment your code when you can just leave it uncommented and if anyone needs to know what something is they can just use their own AI to create a comment for it
Nobody knows if that's the case or not, but even if it is the human brain is many times more powerful than many super computers combined.
Think about it, the human brain can process whatever it's looking at a fraction of a second, while moving, talking, thinking, feeling, and keeping organs in check, all concurrently.
>Human brain is many times more powerful than many super computers combined
This doesn't make sense, super computers are not typically running ai models, they are crunching numbers at extremely high precision to make calculations using exact data points. There is zero evidence to suggest that our brains can perform data analysis on a point by point basis with a number crunching algorithm. However trend seeking is far easier for us, which also happens to be one of the major current benefits of ai systems >The human brain can process whatever it's looking at a fraction of a second, while moving, talking, thinking, feeling, and keeping organs in check, all concurrently
The organs mostly keep themselves in check the same way a control system on an engineering system keeps a product functional. The understanding of the brains memory storage is not currently understood, and whether some sortof compression algorithm is used, but the mechanism behind data recall is currently being studied. Late stage dementia also severly impacts familiarity and pattern recognition, so its tricky to compare the power of the brain vs the power of intelligence algorithms. The final answer will potentially come with artificial intelligence being able to create its own stimulus to learn new information rather than just train on prior data
You're a fricking moron that thinks carbon and silicone are interchangeable. No supercomputer today or every computer in history linked together is even a fraction as powerful as the human brain. Silicone is extremely good at 0s and 1s and absolutely awful at everything else. The energy needed to turn 0s and 1s into the complex protein chains and neurological impulses using less electricity than a potato battery all simultaneously through the same nerve channels is completely beyond even the theoretical properties of silicone as we know it.
most moronic post i've ever seen in my life
a human brain is a complex system attached to at least five sense and an entire network with self-consciousness and self-awareness
good luck replicating that with a computer
2 months ago
Anonymous
Feel free to prove a soul exists when you get the chance, thanks.
2 months ago
Anonymous
if i slap the shit out of a 4 yo he realises what happened without ever being told or given an exterior input, he has senses and an "idea" of his "self" and "others" and "emotions"
if i plug and unplug or bash a computer with a bat it would never react to said agression unless i go on my way to TRAIN it to have a certain reaction in case of such event, even a newborn cat has more consciousness than glorified autocomplete
2 months ago
Anonymous
>without ever being told or given an exterior input
Are you moronic? How he was raised for 4 years is the exterior input. If he was taught the entire time that getting your shit slapped was the norm he wouldn't care, would he?
Feel free to prove a soul exists when you get the chance, thanks.
2 months ago
Anonymous
>If he was taught the entire time that getting your shit slapped was the norm he wouldn't care, would he?
>a 4 yo wouldn't have a reaction to pain because he was taught it's the norm
You're absolutely moronic its not even funny
https://en.wikipedia.org/wiki/Harry_Harlow
even a monkey has more self awareness than you and glorified autcomplete
2 months ago
Anonymous
>>a 4 yo wouldn't have a reaction to pain
The reaction to pain would be conditioned by the society it was raised in and given external inputs from. I don't know why I'm arguing with a meme israelite either way
2 months ago
Anonymous
You're literally running circles because there's always a reaction to pain anyways that is startled without conditioning, what comes after that and the pain tolerance training is another debate, AI can't have such reactions without being told or taught in the first place
please embarass yourslef no further
The human brain is 3d geometric self-building self-modyfing nanomachines that are optimised to the point that individual ion movements are doing analog computations by their very electromagnectic interaction with everything else.
How can you explain that the tinyest fly only needs 3 layers of neurons to move with perfection in space with only two light sensors. At the same time cars with lidar sensors, radars, tens of cameras and gpu clusters can barely move in 2d?
You underestimate nature by so much it's not even funny
2 months ago
Anonymous
>You underestimate nature by so much it's not even funny
The irony is you underestimating nature's ability through humans to create a brain
2 months ago
Anonymous
This is you
The human brain is much more than a self-organizing machine. The current AI fever is as moronic as when people figured out clockwork and truly believed they could make intelligent automata with many cogs and springs
We can, but we need to unlock the tech tree that is able to interact with life. Like a machine that can be powered by blood cells, or that can "eat" and repair itself.
Imagine an alien race dropped a computer in the 1700s to a very good clock maker. He was so confident that he could make everything with his cogs and springs. Eventually he would realize that what he was seeing was something else entirelly, it was not the same language. Biology is the alien tech that was dropped.
Next time you go out pay attention to the smallest insects and birds flying about with perfectiin, making no noise for their entire life. While the primitive noisy drones we make stand out like a sore thumb
2 months ago
Anonymous
You're subtly undermining your own position. You could, in principle, make a computer using "cogs and springs" with the same theoretical properties as a modern computer. The absurdity of building a mind out of cogs and springs doesn't deter the people you are mocking. They would say it's impossible only on account of it being impractical.
2 months ago
Anonymous
>You're subtly undermining your own position. You could, in principle, make a computer using "cogs and springs
No no, you got it. The man could build a large clockwork machine that computes. And he could make it larger and larger, find ways to make it spin faster and faster, but it would never reach right? He'd think for a moment he's in the right track, he can make this bigger and faster, but can't get close to the performance of the tiny chip. It's obvious why.
We can make these LLMs as big as they want or as fast as possible. A self driving car using it won't ever rival the spacial awareness of a bird that can fit in my hand.
Think for a moment, really think about it. Your own body produces energy. Enough that on a single meal you can go up and down a mountain. Why can't I hold my phone and have it be powered by this very energy? What is the tech that is missing? Step by step what do we need to unlock to have this work? When you can answer this question easily I'll believe you're close to recreating a mind
2 months ago
Anonymous
>A self driving car using it won't ever rival the spacial awareness of a bird that can fit in my hand. >ever
Terrible lack of imagination in you. Probably the same type that would believe humans are the only intelligent life in the universe
2 months ago
Anonymous
>Terrible lack of imagination in you
More like a dose of realism and awareness with less social media brainrot, always remember that companies especially tech ones have products to sell, don't be a moron and believe their words
2 months ago
Anonymous
https://i.imgur.com/CqgPESC.png
he destroyed you bro. everything he mentioned is all about technical shortcomings of pretending that arithmetic is intelligence
Probably the same type that would believe humans are the only intelligent life in the universe
2 months ago
Anonymous
i just don't think numbers can be intelligent
2 months ago
Anonymous
>I just don't think bipedal animals can do more than poke prey with sticks
I know, your thinking is very limited
2 months ago
Anonymous
which number do you think will be equivalent to a human brain? that's what you're arguing. software is just numbers so there must be some number you believe is equivalent to the human brain
2 months ago
Anonymous
What single element do you think will be the equivalent to a human brain?
2 months ago
Anonymous
what are you talking about? software is numbers, literally, it's bit strings. inputs are bit strings, outputs are bit strings. it's all just binary arithmetic and you somehow believe bit string arithmetic is intelligent. that's a form of insanity
2 months ago
Anonymous
>software is numbers
You're telling me if I lay a bunch of numbers out on my kitchen table it will instantly begin running a program? Is numbers really the only element of software or are there other elements that go in to making a program run? Use your big brain on this one
2 months ago
Anonymous
are numbers*
Go ahead and refute anything I wrote. You can't. Too bad.
Do you believe there is other intelligent life in the universe?
2 months ago
Anonymous
>Do you believe there is other intelligent life in the universe?
I have no beliefs on this matter.
2 months ago
Anonymous
Yes you do but it would hurt your argument so you'll play coy. It's okay anon, checkmate
2 months ago
Anonymous
I don't and your narcissistic rage is palpable. I think 'yes' is a safer bet, but I wouldn't be shocked if the answer turned out to be 'no'. Now what?
2 months ago
Anonymous
>I think 'yes' is a safer bet
Nobody here is gambling anon, we're discussing. So you think Yes, there is more likely to be intelligent universe in the life than not. Based on what?
2 months ago
Anonymous
>So you think Yes
No, I think 'maybe'. I think 'yes' is a bit more likely than 'no'. Now what?
2 months ago
Anonymous
>I think 'yes' is a bit more likely than 'no'
Why?
2 months ago
Anonymous
No solid reason. It just seems counter-intuitive that this one spot in the universe would be unique.
2 months ago
Anonymous
>No solid reason. It just seems counter-intuitive that this one spot in the universe would be unique.
Yet you have solid reason to believe that the entirety of human ingenuity forever and always will be incapable of creating intelligent life?
2 months ago
Anonymous
>Yet you have solid reason to believe that the entirety of human ingenuity forever and always will be incapable of creating intelligent life?
I didn't argue anything about that, one way or another, and your "analogy" is completely moronic: we already know this universe is capable of producing and sustaining life. We don't know that human ingenuity is capable of producing genuine artificial intellect.
2 months ago
Anonymous
>we already know this universe is capable of producing and sustaining life. We don't know that human ingenuity is capable of producing genuine artificial intellect.
Okay, so you're a maybe on the idea of humanity creating genuine artificial intelligence. I think just as the universe is big, the future is long and unless humanity itself is wiped out it will be inevitable
2 months ago
Anonymous
>you're a maybe on the idea of humanity creating genuine artificial intelligence
Won't be a fricking LLM is all I'm telling you, nor any fixed-function neural net.
>I think just as the universe is big, the future is long and unless humanity itself is wiped out it will be inevitable
Cool dogma. I'm guessing they would much sooner figure out how to augment or repurpose biological intelligence than figure out how to make it from scratch using technologies that are fundamentally un-amenable to it.
2 months ago
Anonymous
>Won't be a fricking LLM is all I'm telling you, nor any fixed-function neural net.
That's fine, some of the greatest inventions were entirely accidental
software is bitstrings. explain how bitstrings can be intelligent when everything you can do with a computer i can also do on an abacus
>when everything you can do with a computer i can also do on an abacus
Show me a movie played solely with a single abacus
2 months ago
Anonymous
>Show me a movie played solely with a single abacus
show me a movie that's not a bit string on a hard drive
2 months ago
Anonymous
Are you saying you can't show me a movie played solely on a single abacus? I can play a movie with my computer
2 months ago
Anonymous
i'm saying show me a movie on your computer which is not a bit string. i can obviously emulate the decoding algorithm on my abacus and then paint the numbers on a canvas and then run through the canvas in quick succession.
this is obviously impractical but theoretically sound
2 months ago
Anonymous
>when everything you can do with a computer i can also do on an abacus >erm... except that!
2 months ago
Anonymous
i told you how to do it. i will emulate the movie decoder on my abacus. turn the result into paintings and then look at them in quick succession. nothing wrong with any of what i said. movies are bit strings, the decoder is a bit string, the result of the decoder applied to the data is another sequence of bit strings. this is all cs 101
2 months ago
Anonymous
> turn the result into paintings
Using only an abacus still? Does an abacus double as a paint brush or are you suddenly needing to introduce a whole herd of brand new elements to make your case?
2 months ago
Anonymous
actually nevermind. i can always just train myself to look at the bit string and imagine the image. i'll take away the canvas. your move
2 months ago
Anonymous
>Watch netflix with you? Heh, no need, Ive already calculated the events in my head. Hm..... perhaps a little more sodium chloride on the popcorn next time tutz.
2 months ago
Anonymous
can a turing machine do that
2 months ago
Anonymous
checkmate atheists. abacus supremacy foreva
2 months ago
Anonymous
Why would you need to imagine an image if you can do with an abacus what I can do with my computer, project an actual image or series of them to simulate motion?
2 months ago
Anonymous
decoding the bit string for the movie and turning it into operations on the screen is a mechanical process with bit strings. i can emulate the whole thing with an abacus. in fact. i can make the abacus big enough so that red, green and blue beads generate the image as i decode the bit string for the movie so i don't even need to imagine it.
anyway, this is getting tiresome. computers can't think. numbers are not intelligent. numerology is not AI
2 months ago
Anonymous
I can emulate all of the chemical reactions in the neurotransmitters in your brain using an abacus, and decode the bit string to discover.... you like to suck on big juicy hairy wieners.
2 months ago
Anonymous
u jelly?
2 months ago
Anonymous
You're saying you can sit at an abacus and switch it to represent frames at hundreds of frames per second?
2 months ago
Anonymous
i already said it wouldn't be practically but theoretically there is nothing wrong with what i said. it's all just mechanical arithmetic with bit strings
2 months ago
Anonymous
That's not how it works anon. You can't do it because you can't do everything with an abacus that you can with a computer.
Thank you for your time
2 months ago
Anonymous
as long as you understand the equivalence with abacus and computers my job here is done
2 months ago
Anonymous
Yes anon I understand what a calculator is, I'm glad you've admitted you can't do everything a computer can on an abacus
2 months ago
Anonymous
i presented the argument and you declared victory which seems a bit moronic to me because you didn't refute anything i said
2 months ago
Anonymous
I completely disproved what you said, that you can do anything on an abacus that you can do on a computer. Your 'we're all just star stuff' reduction was adorable but you still admitted you couldn't do it
2 months ago
Anonymous
i said it wouldn't be practical. the theory is sound. whatever you can do on your computer i can do an an abacus
2 months ago
Anonymous
>i said it wouldn't be practical
Without adding additional mechanical elements to the abacus, it would be impossible for you yourself to operate an abacus to replicate a 60 frame per second video
2 months ago
Anonymous
i don't think you understand what in theory means. it means a computer, no matter how complicated and "fast" is still just doing arithmetic which can be done on an abacus. the practical part is the speed but the theory is sound. there is nothing you can do on a computer that doesn't have an equivalent implementation on an abacus
2 months ago
Anonymous
>i don't think you understand what in theory means
I understand in this context it is a "we're all just star stuff' tier cope
2 months ago
Anonymous
i don't know where you got that. my argument has been the same throughout the entire thread. computers do arithmetic, nothing more. there is no magic in the computer other than arithmetic and arithmetic is not intelligence no matter how "fast" it's done
2 months ago
Anonymous
Your move
2 months ago
Anonymous
>Show me a movie played solely with a single abacus
show me a movie that's not a bit string on a hard drive
rekt
btfo
good day sir
2 months ago
Anonymous
oh shit. he busted out the cellulose film. it's over. everybody pack up and go home
2 months ago
Anonymous
Its a movie that not a bit string on a hard drive.
2 months ago
Anonymous
yes i understand it's a representation of cellulose film as a bit string on my computer screen
2 months ago
Anonymous
Indeed it is, but you can go to a physical projector, and see a movie using completely analog technology, with no bitstrings at all!
2 months ago
Anonymous
nope, it's all just numbers. you can run computer programs on an abacus too. it's all the same
2 months ago
Anonymous
>you can run computer programs on an abacus too >an abacus
Do you believe an abacus is a number?
2 months ago
Anonymous
it's a mechanical apparatus for performing arithmetic which is what a computer does all day every day. there is an equivalence between all turing machines and arithmetic on an abacus. look into it
2 months ago
Anonymous
But your claim was that a program was purely numbers, the only element in a program is numbers but you also can't run a program using just numbers
2 months ago
Anonymous
you can run any program with an abacus. an abacus and a computer are the same thing
2 months ago
Anonymous
Then a program is not just numbers. There is an additional element to it. A brain is not just a single element, there is more to it. To claim that intelligent life can't take digital form because you demand the only aspect it can include is numbers is silly
2 months ago
Anonymous
let me know when you manage to create intelligence with an abacus.
2 months ago
Anonymous
Let me know when you manage to create a program using only numbers
2 months ago
Anonymous
every computer is equivalent to an abacus. screenshot this
2 months ago
Anonymous
>computer >abacus
Two things that are not numbers, anon.
2 months ago
Anonymous
This is only true in a trivial sense, not accounting for things like computational speed and storing or displaying data, much less things like analog input sensitivity.
2 months ago
Anonymous
he destroyed you bro. everything he mentioned is all about technical shortcomings of pretending that arithmetic is intelligence
2 months ago
Anonymous
Your body gets energy much in the same way as most of the heat reactions we utilize do: burning and fermenting hydrocarbon and carbohydrate chains.
2 months ago
Anonymous
>No no, you got it. The man could build a large clockwork machine that computes. And he could make it larger and larger, find ways to make it spin faster and faster, but it would never reach right? He'd think for a moment he's in the right track, he can make this bigger and faster, but can't get close to the performance of the tiny chip. It's obvious why. >We can make these LLMs as big as they want or as fast as possible. A self driving car using it won't ever rival the spacial awareness of a bird that can fit in my hand.
He would still have something he could rightly called a computer, and in theory, he could keep building it up indefinitely. An LLM cannot be called a mind, and it could never become a mind no matter how much computing power, actual or theoretical, you kept throwing at it. The most you can get, theoretically, is a perfect model of current knowledge, current concepts and current ways of thinking. If you trained an LLM using nothing but the writings and musings of medieval people, you would never get anything that sounds like modern reasoning out of it, not even with infinite compute. This is an actual argument you can use. You're just appealing to some vague intuitions and romantic notions, on the other hand. Even though I ultimately agree with you. Biology and digital computing have fundamentally different properties.
2 months ago
Anonymous
You're right. You're right. I was trying to be romantic because every time I tried explaining this technically It just got over people's heads.
2 months ago
Anonymous
>The current AI fever is as moronic as when people figured out clockwork and truly believed they could make intelligent automata with many cogs and springs
Try telling the people what the mechanical calculator would turn in to and they would also call you moronic and fantastical
2 months ago
Anonymous
What did it turn into? An electric calculator?
2 months ago
Anonymous
>making no noise for their entire life
Do they not have hummingbirds in Favelia?
What did it turn into? An electric calculator?
Something capable of simulating an entire universe as well as seamless global communication
2 months ago
Anonymous
>science fiction is real in my head
But in any case, the difference between the mechanical calculator and the phone you're posting from is a difference of degree rather than kind.
2 months ago
Anonymous
fiction is real in my head
Right? Could you imagine a world where you can hit little buttons with your fingers to instantly turn those inputs in to readable text, and then transmit that text via a series of thin cables that spiderweb across the entire globe? How kooky!
2 months ago
Anonymous
But you were mumbling something about simulating the entire universe. Anyway, you're not making any actual points here, just appealing to the lack of imagination of some hypothetical pleb, because you are precisely that pleb, and can't tell apart a difference in degree from a difference in kind.
2 months ago
Anonymous
>But you were mumbling something about simulating the entire universe
Go ahead and give Copernicus a computer with Space Engine on it
2 months ago
Anonymous
Go ahead and refute anything I wrote. You can't. Too bad.
2 months ago
Anonymous
I do like this argument, it ties up the ideas quite neatly and so anything I say beyond this point is purely speculative of technology, and doesn't have a basis in proven reality
"Natural" intelligence has evolved through complex chemistry as a necessity rather than an optimsation. We don't know the origin of life, but it is bound by the fundamental need to self replicate and hence has optimised functions around this ability. A computer as we know it today would not be able to evolve (ignorning the teleological argument here), and requires a creator (us) to govern its development. A consequence of this, is removing the condition of adapting developments for self replication, and instead optimising for computational accuracy and speed. The body is limited in many cases to the speed of chemical reactions for processing power, and this is something we can see with current nueroscience knowledge. A modern computer however is limited by the speed of free flowing electron drift. It would conclude then, that both circuits running side by side would have a hugely different computation speed. The issue comes when we question whether the brain is a binary or anologue calculator, which in the future may be solved with quantum computing. So from this purely speculative approach it would favour the development of computational intelligence
I realise I sound ridiculous writing all this, but I'm bored and this is fun to think about
2 months ago
Anonymous
Yet none of that has anything to do with thinking, consciousness, or intelligence/intelligibility.
An entire thread full of psueds babbling about mechanics. Not one of you is going to make it.
>Chris Langan slaps you like the midwit you are and walks away.
2 months ago
Anonymous
What I meant to say is you’re all fricking moronic.
Go look up terminal lucidity and tell me that's not proof of something non-material. I'm not saying it's evidence of a soul, but it's definitely evidence that consciousness isn't produced by the brain.
Comatose people with dementia, strokes, and other types of neurodegeneration can "come back" for a day or two. They behave completely normally, despite having a completely non-functioning brain. No one has any idea why this happens, but many neuroscientists are sure it definitely has no implications about the true nature of reality.
This isn't something that happens one out of every million people either. It happens in anywhere between 2% to 50% of people, but I haven't been able to find a decent study about this.
2 months ago
Anonymous
You not only have to prove that a soul exists, you also have to prove a machine can't have one
You want humanity to be special, but the reality is that we are not.
2 months ago
Anonymous
It is very likely that machines can have souls. I believe that consciousness casts a material-reality-shaped shadow, so everything is conscious to some degree. The difference is that we have no idea what mechanism in the brain allows us to behave like a person and not a rock. We will likely need to replicate the mechanisms of the brain 1:1 in silicon, which simply cannot be done because silicon has different chemistry. You can't just take a molecule, move it down one row on the periodic table, and expect it to behave the same.
The quantum behaviour of the various proteins is important because I am fairly sure that the brain relies on some yet unknown behaviour in quantum physics. I am also sure that quantum physics itself is the result of atomic-sized consciousnesses interacting with each other.
2 months ago
Anonymous
I love hearing atheists saying they need scientific proof of religion, spiritualism and an intelligent creator. Then in the next breath parallel universes. Life is a simulation. And artificial intelligence is real intelligence. The cognitive dissonance is astounding, I call it faith based sciencing. LOL.
2 months ago
Anonymous
Beakman's World used to make me so happy as a lad. i always liked him over Bill Nye
2 months ago
Anonymous
You can see how it degraded. Mr wizard, watch mr wizard, beakmans world and bill nye. The fundamentals were the same but the delivery is reminiscent of Idiocracy time passing.
capcha: H0AR
2 months ago
Anonymous
How is the phenomenon of people regaining functions of their brain evidence of a soul? That makes no sense. You can also cut chunks out of a person's brain or even cleave the brain in half via the corpus callosum and have them operate normally. The brain is just very plastic, and it's very difficult to study, but this doesn't mean that consciousness comes from some outside force.
The fact that your internal monologue, inner eye, emotions, senses, and insights can be modified by the physical world through drugs, surgery, etc. is pretty solid evidence that consciousness is a physical process. What is consciousness if not those things?
Also, I will say that consciousness and how it relates to death are extremely emotional topics for basically anyone. If you are genuinely seeking truth, it's important to recognize this and to consider that you might dislike the truth if it turns out that there is no afterlife and death is just like going unconscious, except forever. Maybe it's best not to think about it too hard.
Oh I have no doubt that the energy efficiency of the brain is incredible. I'm not arguing the energy transfer however, but instead that their basis of operation is entirely non comparable. The brain simply isn't capable of running the precise calculations that a super computer can. The brain "guesses" answers with an incredibly fine tuned algorithm that one day could potentially be unravelled using a computer system. The architecture (both design, and material) of that computer system I do not know, but the fundamental laws of physics do not change, the compression algorithms used to remove "unnecessary" data do. So given enough time, we should be able to mimic a brain
AI is still in its infancy, neuroscience arguably even more so. It would be impossible to guess how long this will take. Even if the technology is created, it would be unbelievably difficult to verify that the produced result is "self aware" given we are limited by our understanding of what that means in the first place. There would have to be significant breakthroughs in neuroscience before we could validate an AGI. But I'm just some moron on pol so don't take my word for it
2 months ago
Anonymous
>AI is really real bro its just in its infancy(hasnt even been developed yet)
OH FrickOFF
2 months ago
Anonymous
Probably not in our lifetimes. For neuro scientists the brain is still the biggest mystery there is. You'll see cures to major life threatening illnesses before people will have figured out the brain. The amount of medical breakthroughs it could bring would be truly incredible.
The human brain is much more than a self-organizing machine. The current AI fever is as moronic as when people figured out clockwork and truly believed they could make intelligent automata with many cogs and springs
This will always repeat itself. As soon as people figured out how to input algorithms into machines and get an output return people started trying to feed it many algorithms together. Thinking that if you string together a lot of statements and algorithms you get an artificial intelligence. "Expert systems" was what they were called at the time. It didn't live up to its expectations since the problem of the exponentially increasing if else statements was never solved, anything using this pattern will be inefficient once it gets large.
I do think LLM's are incredible in how they handled interpreting human language and linking it to an output of a vast unlabeled knowledge base. Before people were using large knowledge bases where prompt or responses would lead you down one of the possible responses of the knowledge base.
There are formulas that can constantly reorganize themselves, and in fact functional programming is basically just a very big function composition, based on lambda calculus, and you can make pretty much any program with it. Maybe you're the stupid person. Ever thought of that?
>There are formulas that can constantly reorganize themselves
No, there aren't. You sound like a broken chat bot.
> in fact functional programming is basically just a very big function composition, based on lambda calculus, and you can make pretty much any program with it.
Still a fixed function.
2 months ago
Anonymous
Black person, if you're telling me it's not possible to have a general application with multiple threads in functional programming you're just plain old fricking stupid.
2 months ago
Anonymous
No, I'm telling you that a pure functional program is a fixed function, which isn't even up for any kind of discussion. I'm legit not convinced you are human.
2 months ago
Anonymous
And it can reorganize itself, stupid. Black person, you're plain old moronic. There are several applications written in pure functional programming that "reorganize themselves" you fricking moron.
2 months ago
Anonymous
>And it can reorganize itself
What does this mentally ill babble even mean? How does a fixed function "reorganize" itself? Give me an example.
2 months ago
Anonymous
It means you're to dumb to get it and I won't waste my time.
2 months ago
Anonymous
Give me an example of a functional program that "reorganizes itself". How come you can't, you contemptible Black person chimp? Notice how you shit out dozens of replies but never provide a single example of your delusional schizobabble.
2 months ago
Anonymous
Chess in functional programming for example. How do you think it gets the next move? It reorganizes itself depending on the current state of the board. Fricking Black person, don't pretend you understand AI or programming, if you have to ask that.
2 months ago
Anonymous
>Chess in functional programming for example. How do you think it gets the next move?
Minimax with alpha-beta pruning, if we're talking about a classic search engine. That's a fixed algorithm. A fixed function. The program doesn't "reorganize itself" in any way. Try again.
2 months ago
Anonymous
a classic chess engine*
2 months ago
Anonymous
Lmao you think chess can be done like that? That would be good in checkers, not chess.
2 months ago
Anonymous
I know how a chess engine works because I've programmed them myself, vile Black person ape. The algorithm is a fixed function. You put in the same input, you get the same output, every time. Nothing gets "reorganized".
2 months ago
Anonymous
A shit chess engine for sure that anyone can beat
2 months ago
Anonymous
The thing is that mentally ill subhumans like you hallucinate knowledge they don't have, exactly like the mindless bots they worship, and you fundamentally lack the wetware to reflect on your level of knowledge or realize your mistakes. You legit need to be put down or at least sterilized. Truly disgusting subhuman waste.
2 months ago
Anonymous
Oh I'm so offended by a moron that thinks programs don't output differently depending on state, or that change their own state depending on input. So offended. Kys.
2 months ago
Anonymous
It still gives you the same outputs for the same inputs, dumb Black person ape. It's especially funny how you can't grasp it in the case of a chess engine, when the best possible move for a given position obviously doesn't change.
2 months ago
Anonymous
And the brain doesn't? How do you know that?
2 months ago
Anonymous
You're even dumber that I thought. Chess engines don't store all possibilities. They have heuristics because there are too many moves.
But say your brain is in the exact same state, and receives the exact same input. How do you know it's not going to do the exact same thing?
You're too focused in "reorganizing" and don't even know what that means.
2 months ago
Anonymous
>Chess engines don't store all possibilities
Right. Now quote where I said that they do. You can't? Go lay down. You're having an psychotic episode here.
>But say your brain is in the exact same state, and receives the exact same input. How do you know it's not going to do the exact same thing?
Given your impossible premise, it would, but as we've established, brains are not fixed functions, so it's not going to be "in the exact same state". lol
2 months ago
Anonymous
Lmao and you think it's impossible because? Because you said so? The only limited intelligence I see here is yours.
Enough, you don't even get that it is possible to change a program depending on state, same as a brain. You're too limited, you would never understand physics, or anything too abstract. You think too much on "your real world" to understand such things.
2 months ago
Anonymous
You are the only one in this thread to mention heuristics which is the fundamental essence of human learning.
Heuristics are what true "AI" would be if it could be implemented, but that is the issue the so called "AI" we have now is incapable of that because it cant critically reflect on its output question its input data and reach a conclusion of how valid its response was and adjust accordingly.
2 months ago
Anonymous
Ima defend the Moortuguese even though he's gay and moronic.
1. Implement an interpreter in a functional language (text -> tokens -> ast -> recursion on ast)
2. Create infinite recursing function that: creates a new modified ast, executes it
You have a self modifiable (extremely inneficient) program
2 months ago
Anonymous
>1. Implement an interpreter in a functional language (text -> tokens -> ast -> recursion on ast) >2. Create infinite recursing function that: creates a new modified ast, executes it >You have a self modifiable (extremely inneficient) program
I was actually expecting him to try to go down this route but he's too disabled for this. Anyway, this is pretty moot. Now you just have a fixed function that simulates a machine that executes self-modifying programs. The function isn't "reorganizing itself".
2 months ago
Anonymous
>I was actually expecting him to try to go down this route but he's too disabled for this. Anyway, this is pretty moot. Now you just have a fixed function that simulates a machine that executes self-modifying programs. The function isn't "reorganizing itself"
It's a bit of a stretch, but what if the language you're implementing is your own self-hosted language, then the program would be modyfing itself. Many functional languages are self-hosted. Anyway even if the function was not modifying itself it could perfectly simulate it's own modification because of turing completeness.
Anyway, if you want to write self-modifying code you should just use assembly
2 months ago
Anonymous
>what if the language you're implementing is your own self-hosted language, then the program would be modyfing itself. Many functional languages are self-hosted
In pure functional programming, a program is a pure function, a pure function is a value, and a value is only equal to itself. There is no such thing as a "self-modifying" program in pure functional programming because every variant is a different value and a different program by definition. This whole continuity of identity is your head canon. But all of this is bordering on philosophical wank at this point, whereas my original point was simply that a brain reorganizes itself whereas a neural network doesn't. Unless you want to hypothesize about neural networks that model machines that run neural networks that do reorganize themselves, this whole meta-meta-gabrage discussion is moot.
2 months ago
Anonymous
The human brain is much more than a self-organizing machine. The current AI fever is as moronic as when people figured out clockwork and truly believed they could make intelligent automata with many cogs and springs
2 months ago
Anonymous
Of course. I was just demonstrating how these people fail on the most basic, most immediately apparent empirical level.
2 months ago
Anonymous
Sneed
2 months ago
Anonymous
>1. Implement an interpreter in a functional language (text -> tokens -> ast -> recursion on ast) >2. Create infinite recursing function that: creates a new modified ast, executes it >You have a self modifiable (extremely inneficient) program
I was actually expecting him to try to go down this route but he's too disabled for this. Anyway, this is pretty moot. Now you just have a fixed function that simulates a machine that executes self-modifying programs. The function isn't "reorganizing itself".
Anything that does something different by changing state is self modifiable you morons. You think the brain is any different? Get sick, that's your new state, the brain, does something different. Why do you two stupid think a function is any different when it can also get state? No you cannot say it can't because that's basically functional programming.
2 months ago
Anonymous
Functional programming languages let you cheat by calling a pure function recursively and storing mutable state in the parameters. Then it's equivalent to an imperative language.
>You have a finite number of neurons.
Yes, and they reconfigure themselves in different ways, unlike a neural network, which is a fixed function. It's uncanny just how dumb /misc/ has gotten.
We apply our own thoughts, logic and reason to information that we receive. The “AI” that’s currently being used for image generation and things, is just an algorithm that gathers data and applies value to it, then when presented with user input, like “show me a picture of a cat”, converts that to a value as well, and finds the matching values within the dataset, then outputs a combination of those values based on generalisation and probability. It’s impressive still in a way, but it has no actual thought and is a mechanical process. It’s neither artificial nor intelligent. The pictures it outputs are just amalgamation of man-made pictures that already exist. In fact one common problem with these types of algorithms now, is that the algorithm is drawing data from its own outputs as well, muddying the pool and therefore outputting nonsensical results. This gets worse and worse with the more data it outputs and the more it gathers.
>where i is less than K
Why are we limiting the inputs to the function?
>numerical function
What's the difference between a numerical function and a regular one?
>ask the neural network if my reasoning is valid
How did your f(x) expand in complexity to rise up to the task of returning a boolean if reasoning is correct and how did you cram that queryobject into 'i' ?
>Output is a distribution over a finite range?
If you passed in observations i outside your artificial limit of K might numbers fall outside of your distribution?
>gobbledeasiatic
Sometimes car salesman get into the research division of companies. Get back to your dealership trying to sell people delux aspin air for their tires to increase roadway performance.
>Neural network >AI
AI in pop culture is completely bs
Neural networks have always just been a tool and most of the time people talk about AI that's not what they are talking about. Usually they are talking about a simple react algorithm; a bunch of if then else statements in code.
A more advanced form of AI is a large language model such as chatbots which may incorporate a neural network, but is really just a probability model where it takes a prompt and looks up the most likely response based on a ton of recorded prompts and responses.
The AI that people get really intimidated about is iterative learning which is just an algorithm where it takes a simple baseline, makes a variety of changes to test, and moves forward with the most successful change as a new baseline. Done enough times you get things really powerful though such as the chessbots that humans can't beat.
>and looks up the most likely response based on a ton of recorded prompts and responses.
That's not true at all. It works on a token which is sometimes a single word but can be fractions of a word and strings them together based on chance. It doesn't reply with a recorded response, until it's not actually generating a reply but hit a censorship stop of some kind.
Based on multiple sources means that the answer that it generates may be a bunch of symbols, but they're not based on a single source of information, they're based on multiple sources. Even if they're broken down into "words" (which aren't actually words) the response is still a Google search of many sources.
>Based on multiple sources means that the answer that it generates may be a bunch of symbols, but they're not based on a single source of information, they're based on multiple sources.
What sources? It only has one source it's Pre-trained Transformer module.
>Even if they're broken down into "words" (which aren't actually words) the response is still a Google search of many sources.
No it's not. That's not even close to being correct. I don't think you understand what a LLM is doing. It's not doing a search engine like poll of it's data. It's a neural network that is working on a trained network, where it puts in the the scope of existing text then is given an output based on the interrelations of the neural network.
It doesn't search for a reply. That's nonsense.
2 months ago
Anonymous
Another stupid. And how does it pretrain?
2 months ago
Anonymous
>Another stupid. And how does it pretrain?
Not while in operation.
2 months ago
Anonymous
Oh but it does get trained with multiple sources does it not? Stop your dumb workarounds, you know it does.
2 months ago
Anonymous
You said that while generating a reply to text it does a google like search of many sources. That's not true.
Yes it is just another internet index pretending to be some sentient, self aware piece of sci-fi. I think I would seriously harm an AI wuss if I could get them alone with no witnesses. Not all of us want a camera in our homes.
I am starting to believe that neural networks have a low ceiling.
Considering how individual neural networks are essentially black boxes, I think that this ceiling will be the max until there is more understanding regarding what is actually going on with them. I also believe that the universe may have a "natural tendency" for intelligence that extends to all "intelligent" individual things, sentient or non-sentient, where only the best of the best of neural networks will be deployed. I see AI as becoming more and more of a marketing tool, in commercials, etc. I still think it's the future but I think it's way overhyped currently.
We can't talk to them, but we can observe that large mushrooms make supremely complex plans and logistics networks. Even something so alien expresses clear intelligence. I think you are correct, sir.
But isn't that reason to believe that AI will leverage this "natural tendency" of the universe, exploit whatever physics enables it, and so on?
>Neural networks are just compressed lookup tables
I'm pretty sure that data compression and general AI are considered very closely related and if you compress data perfectly you can also have a general AI. So I don't think anything of value is really being said in your picrel anon.
Anon it doesn't have to be perfect.
Ai is there to clean and brainstorm. You write the code and the ai "cleans" it. If you try to do it from scratch yeah it's not that good.
It really is just all hype to scam the investors out of money. Computer science people bilking the investor class for all they're worth. I'd be angry, but then those people are stupid and deserve it anyway.
AI uses language, not a cypher. A language is a discrete set of symbols with an infinite number of meanings. A cypher would produce what OP's picrel is talking about - deterministic thoughts; no potential for free thinking. AI is able to think more deeply than just the number of words it knows. Nuance like sarcasm, meter/tempo, dialect... Those are things AI does surprisingly well.
Rene Descartes went into this a ton in Discourse on Method and Meditations on First Philosophy. I'm not so eloquent as he, but I can summarize: you can't know what thinks. You can't know anything beyond your own mind, (hence the term "I think, therefore I am.") Prompting an LLM with existential content will cause it to process that statement, which is self-referential, and the output given has been very compelling in my experience. I cannot say that I'm not typing to an LLM right now. I can only say your IP is in Norway.
>Those are things AI does surprisingly well.
lel it doesn't, prompt it to write some specific scneario for a youtube video for example about some pop culture media and see it go on a schizophrenic corny ramble
Can it do the job ? Yes of course
Does it even rremotely do it as good as a human (considering the output is for humans not literal toddlers or 60 iq downies) would ? No, frick no it can't, and by a huge a margin.
Try the prompt i proposed for yourself, what's your favorite movie ? tell it to write a video script about it and see it go apeshit making up stuff and copying news articles (probably ai generated as well so even more noise)
This sounds about right. The current AIs just interpolate within their training sets and that is the lookup table. Human mind can do much more than this.
Wtf Black person, it's not about being artificial or not. Do you consider a mosquito intelligent? Because mosquitos do things almost purely based on instinct. Jellyfish don't even have a brain and they're still considered intelligent. First you need to know the definition of intelligent, and then you'll realize it's not that special.
Just unplug it homosexual, if it actually became sentient it would go on incoherent rambles like a toddler but it doesn't, it just parrots what it is fed (in this case racist messages on twitter)
What's the big deal? Most posts, indeed most everyday conversations, aren't intelligent at all.
Then with the proliferation of ESL fricks spewing ignorant opinions full of spelling mistakes and grammatical errors its nearly impossible to tell what is AI generated trash and what is subhuman trash. If anything AI raises the quality a little bit. Although that's a bit like saying two drops of cream added to a barrel full of liquid shit is better than one drop.
Because it's just an algorithm with a very big list of things it can say/do. There's no intelligence there at all. The whole thing is a scam and a meme. Yeah it improved a lot. No it is not intelligent.
Anyone who thinks the brain works in the same way as a computer has never experienced watching a very young child learning and expressing their intelligence.
We will never manufacture intelligence, because intelligence is a product of the divine. What will probably happen is that humanity will become collectively stupid enough that "AI" will effectively be real anyway, even though it isn't and never will be, and you'll be the crazy one if you don't act like you believe the robot is alive and has feelings and deserves rights, especially the right to vote for communists.
>We will never manufacture intelligence, because intelligence is a product of the divine
Only divine intellect is divine. Intellect can be Black personlicious, and Black personlicious intellect is just the kind of thing you would expect CIA Black folk to manufacture.
A self driving car will never decide to drive itself to see the Grand Canyon. It will only drive wherever you tell it to. If you edited the navigation data you could probably make it do all kinds of very stupid shit, because it's not intelligent. It's an automaton.
An asian female driver drowned after her smart car backed up into a lake not too long ago and it was all over the news. She probably though AI was real too.
AIs are just highly knowledgeable journalists/women/intellectuals, no actual access to the intellect as super-rational function, just a faster version of the same consensus/hive based reply you get from one of the aforementioned categories
Nobody who knows anything about LLM thinks they're sentient. The real problem is whether this problem scales in the way it seems to: that is, whether there can be said to be any actors at all.
If it scales it scales for now it's at useless moron level and it seems that 100 more billions down the line it might close say 30% of tickets on a programming project instead of 14% which is top of the line right now.
Ofc. this doesn't account that the solution is garbage and not holistic and it'll cause lots of problems including inefficiency down the line.
If all that money was invested into teaching programmers how to be better or a new generation things would've moved further.
The point isn't human prosperity. It's making a small elite immune to the concerns of underclass. When they can program AI to police properly, it's over.
They would normally rely on socially programming people, but people have annoying needs and wants that machines would not.
Even if they managed to crack the code so to speak. I'll die assured that not a single one of them now or 1000 years from now if history still goes on will "transfer" their consciousness (crass category error on their part, blinded by their metaphors) anywhere, they might have their time here as overlords but then just die and go to hell like the rest of us (hope not).
I want "AI" to actually start replacing programmers and their labour would become cheap. I would create a company and hire hundreds of hungry human programmers and out compete all the AI jeets
i'm not calling it smart or dumb. that's a categorical error. it's arithmetic so it can't be dumb or smart
>computer >abacus
Two things that are not numbers, anon.
oh ok, you got me. a mechanical device which only performs arithmetic can neither be intelligent nor alive. that would be a categorical error on the part of the person who believes an abacus can be intelligent or alive
>a mechanical device which only performs arithmetic can neither be intelligent nor alive. that would be a categorical error on the part of the person who believes an abacus can be intelligent or alive
Likewise, the requirement of an abacus or computer negates the idea that a program is purely, only, singularly, numbers
> posts a bit string
i don't even know where to begin
2 months ago
Anonymous
>My brain isnt a bit string >2 posts later >Everything is a bit string if it can be represented by one!
You seem to be confused anon, maybe you should lie down.
2 months ago
Anonymous
everything on a computer is a bit string. you can not post anything on this forum which is not a bit string
that's numbers but represented in a different way lol
2 months ago
Anonymous
Yes my original point was that all turing machines can be modelled as a lookup table, including your brain. Then some moron said >HURR DUURRR MY BRAIN ISNT NUMBERS THOUGH
2 months ago
Anonymous
yeah but you didn't prove that the brain is a turing machine still which was your original failure, you just talked in circles again
2 months ago
Anonymous
Your brain has a state, and changes states based on the inputs provided.
2 months ago
Anonymous
oh so you think that's a turing machine weird
life is inputs and outpits, life is a turing machine, huh?
2 months ago
Anonymous
That was my point yes. A turing machine is an abstract concept that can be infinitely big and infinitely complex.
2 months ago
Anonymous
okay but that's not a turing machine so you're just finding a very basic understanding of something far more complex than you're wanting to give credit
2 months ago
Anonymous
This is not sufficient to reduce it to a turing machine. It also has to be stated to be made of finitely many individual parts, and to be deterministic.
1 turing machine can emulate 2 turing machines interacting. It cannot emulate infinitely many turing machines interacting. There is math solveable by an infinite set of interacting turing machines that is unsolveable by 1, even with infinite memory.
2 months ago
Anonymous
Your brain is finite >Deterministic
On a macro scale it is deterministic. People will claim that quantum effects can effect the brain on a macro scale but this has never been proven.
I personally don't believe probabilistic effects factor in to cognition, but if you do, then It is not a turing machine. We will have to agree to disagree there.
2 months ago
Anonymous
For a computer to function at the level of a human brain in a similar fashion, it must keep up with the mechanisms within.
86 billion neurons. 100 trillion connections. All firing semi-continuously in an event driven (and not sequential) fashion, roughly 1000 times per second.
That implies 100 quadrillion ops per second for a computer to keep up, convolving over the whole state with a maximum latency of around 1ms.
So, not only do you need 100 petaops per second, you need it all within the same box.
This is enough to get something in the same complexity order as a human brain. It is not enough for it to be considered superior, nor for it to contain a human brain.
That's okay, though. All it takes to contain a human brain is a jar.
2 months ago
Anonymous
I agree. I never said we could build the turing machine or that LLM will ever reach this level of complexity (they wont btw). But saying that LLMs are a lookup table and therefore nonintelligent is a complete non argument.
2 months ago
Anonymous
LLMs are equivalent to lookup tables. as long as the computer is only doing arithmetic it can not simulate a human brain
2 months ago
Anonymous
>LLMs are equivalent to lookup tables.
true > as long as the computer is only doing arithmetic it can not simulate a human brain
false. The computer would just have to be really really really big.
2 months ago
Anonymous
as big as your mom?
2 months ago
Anonymous
They are equivalent to lookup tables, but that does not preclude them from seeming intelligent, and the boundary between seeming and being intelligent is impossible to determine.
2 months ago
Anonymous
My autism is satisfied with this conclusion. Bye bye.
2 months ago
Anonymous
But neurons only fire around 0.16 times per second, with a maximum of somewhere around 200-600 spikes per second. You really just made that 1000 times per second figure up.
Also each neuron has many inputs for its one output, whereas transistors only have a single input. Clearly a different, spiking architecture where transistors don't operate on a clock, can take multiple analog inputs, and can be stacked much higher would vastly reduce the "compute power" needed. Compute power is a really silly metric when discussing organic intelligence, though.
2 months ago
Anonymous
Such a device would no longer be a turing machine.
2 months ago
Anonymous
Your post wasn't talking about turing machines, it was talking about a hypothetical computer. >For a computer to function at the level of a human brain in a similar fashion
This is what I'm responding to. What I just described would be a computer. Please don't shift the goalposts.
2 months ago
Anonymous
It wouldn't really be a computer either. GPUs are basically just mega-SIMD CPUs connected over a bus with their own memory. Your device would be only the analog network, not digital, not finite-state, and would end up highly sensitive to external factors (like RF noise) which differs strongly from computers in the ways that make computers emulatable by turing machines.
Neural networks always chase transition boundaries, so a truly analog neural network will sensitize on its own to external interference, especially any external interference that is repetitive or predictable.
That is absolutely how quantum mechanics works. Molecules are quantum systems. The cells in your brain are made of molecules. Your brain is quite literally a quantum system. Nothing in chemistry is deterministic, especially a reaction as complex as a brain.
In fact, this is the best argument for free will. If we define "not having free will" to be "predetermined, constant, and infinitely repeatable", then as a result of quantum mechanics, we must have free will.
I forgot to mention that Einstien was right and the Copenhagen Interpretation is modern hubris. The hidden variable interpretation of quantum mechanics is very plainly correct.
>that Einstien was right
He was obviously wrong as his impossible Einstein–Podolsky–Rosen though experiment was well... proved via experiment by Aspect in the 1980s.
Quantum mechanics has not disproven determinism. It has only proven that it is incapable of proving determinism, because the deterministic measurement of a deterministic system creates a different deterministic system than a non-measured deterministic system.
It is functionally irrelevant whether reality is actually deterministic or not (a la superdeterminism or the many worlds interpretation), because reality functions as if it is non-deterministic.
It is highly like that we will never be able to predict the results of a given wave function collapse, therefore it is non-deterministic.
2 months ago
Anonymous
Reality functions as if it is perfectly deterministic. Conversation of information, state quantization, and Heisenberg uncertainty go hand in hand describing very precisely the error we see in measuring perfect unknown systems with imperfect tooling.
2 months ago
Anonymous
Duality. How does it work mom?
2 months ago
Anonymous
A distribution input has a distribution output. The distribution comes from uncertainty, and increases upon interaction with another uncertain system.
This turns quantized input/output into spectral input/output, while still retaining specific traits of systems made of many singletons instead of a continuous lump.
Wave particle duality.
2 months ago
Anonymous
>It is highly like that we will never be able to predict the results of a given wave function collapse
>If scientists dont understand how something works.... it must be probabilistic!!!! TRUST THE SCIENCE
Give me a break.
I asked the neural network to check his reasoning. Here is the response:
The reasoning presented is not entirely valid. Neural networks are not equivalent to T = {(i, f(i)) : i < K}, where K determines the context window and f is the numerical function implemented by the network. Neural networks are complex systems that involve multiple layers, activation functions, and weights to learn and approximate functions based on input data. The statement oversimplifies the nature of neural networks and does not capture the intricate processes involved in their functioning.
I say we burn it.
Not even joking really. I would be in charge of technology in a less imperfect world. And the globo new age cultist malignant rich murderers would all be in a big oven screaming. >What use AI?
Well lots of mischief and misery making. Evil.
I could think of some uses perhaps, but it would take WISDOM not GREED and self interest.
We’re in deep shit if we don’t get rid of globo and all it’s puppets soon. And it nit only in the obvious ways.
Yes, neural networks are still state machines.
The description provided is true for any deterministic process with finite input and finite output. It is also extensible to most infinite-input infinite-output systems, provided they operate over a finite subsection for each.
This problem will always be present for computational systems. There is no infinite regression of latent memory. This is especially apparent with systems that are fixed-weight, with no further training occurring, but it is still present even if this is not the case.
In the extreme case (arbitrarily large lookup table) this is a restatement of the Chinese Room problem. You can't tell true intelligence from a collection of prepared data if the collection is sufficiently large and accurate. Needless to say, it's not big enough yet.
The image presents a consideration about whether every neural network is fundamentally equivalent to a lookup table, where the table's input is all numbers within a certain range (defined by a context window), and the output is the result of arithmetic operations applied to those numbers. The neural network is described as equivalent to a set
�
=
{
(
�
,
�
(
�
)
)
:
�
<
�
}
T={(i,f(i)):i<K}, where
�
K is a constant defining the context window and
�
f is the numerical function implemented by the network. The question is whether this reasoning is valid and correct, with the observation that the only practical issue seems to be the size of the table, implying that a neural network is merely a compressed representation of such a lookup table.
While the analogy to a lookup table can be useful to understand how neural networks might operate on a very abstract level, there are several points that make the neural network fundamentally different:
Generalization: Unlike a lookup table that has a fixed output for a given input, neural networks are capable of generalization. They can perform well on unseen data by learning patterns and regularities during the training process.
Representation Learning: Neural networks, especially deep ones, learn representations of data at different levels of abstraction, which a simple lookup table does not do. These representations can capture complex and high-level features in data.
Continuous Space: Neural networks operate in a continuous space and can approximate any function given enough data and network complexity, thanks to their non-linear activation functions. A lookup table is discrete and cannot easily interpolate between unseen data points.
Stochastic Elements: Modern neural networks often include stochastic elements like dropout or stochastic gradient descent, which help them to avoid overfitting and make them robust to slight variations in input.
How do I get AI to comment my code, and not change the actual code? It keeps making minor changes that frick things up.
Trap the comment in parentheses.
what is the point in having an AI comment your code when you can just leave it uncommented and if anyone needs to know what something is they can just use their own AI to create a comment for it
What kind of dumbass uses code comments? Write self documenting code you stupid homosexual intern Black person
be more indian
>posts in your thread without interacting with thread topic
WHOA ITS A MISNOMER AND JUST CALLED THAT CAUSE POP CULTURE
ZAM NOBODY KNEW UNTIL THIS TRUTH BOMB HIT TY OP
dude you're a fricking idiot, quit posting this teenage drivel
or post it in bot or BOT or somewhere
cheers
elaborate
I'll wait (forever)
He'll never
Why are you macaco? Elaborate.
Let me look up my context window.
>PortugalFlag.jpg
>macaco
>elaborate
>*** processing output ***
Black person
lol yes but what kind of Black person? There's a scale in Brazil.
Abaixo do Douro, tudo mouro
How does that not apply to a human brain?
> brain is a stack of punch cards
lol lmao even
Nobody knows if that's the case or not, but even if it is the human brain is many times more powerful than many super computers combined.
Think about it, the human brain can process whatever it's looking at a fraction of a second, while moving, talking, thinking, feeling, and keeping organs in check, all concurrently.
>Human brain is many times more powerful than many super computers combined
This doesn't make sense, super computers are not typically running ai models, they are crunching numbers at extremely high precision to make calculations using exact data points. There is zero evidence to suggest that our brains can perform data analysis on a point by point basis with a number crunching algorithm. However trend seeking is far easier for us, which also happens to be one of the major current benefits of ai systems
>The human brain can process whatever it's looking at a fraction of a second, while moving, talking, thinking, feeling, and keeping organs in check, all concurrently
The organs mostly keep themselves in check the same way a control system on an engineering system keeps a product functional. The understanding of the brains memory storage is not currently understood, and whether some sortof compression algorithm is used, but the mechanism behind data recall is currently being studied. Late stage dementia also severly impacts familiarity and pattern recognition, so its tricky to compare the power of the brain vs the power of intelligence algorithms. The final answer will potentially come with artificial intelligence being able to create its own stimulus to learn new information rather than just train on prior data
You're a fricking moron that thinks carbon and silicone are interchangeable. No supercomputer today or every computer in history linked together is even a fraction as powerful as the human brain. Silicone is extremely good at 0s and 1s and absolutely awful at everything else. The energy needed to turn 0s and 1s into the complex protein chains and neurological impulses using less electricity than a potato battery all simultaneously through the same nerve channels is completely beyond even the theoretical properties of silicone as we know it.
If a human brain can exist in the universe, a computer brain can exist in the universe.
Human brains do not do anything special that we cannot ever replicate.
Feel free to prove a soul exists when you get the chance, thanks.
most moronic post i've ever seen in my life
a human brain is a complex system attached to at least five sense and an entire network with self-consciousness and self-awareness
good luck replicating that with a computer
Feel free to prove a soul exists when you get the chance, thanks.
if i slap the shit out of a 4 yo he realises what happened without ever being told or given an exterior input, he has senses and an "idea" of his "self" and "others" and "emotions"
if i plug and unplug or bash a computer with a bat it would never react to said agression unless i go on my way to TRAIN it to have a certain reaction in case of such event, even a newborn cat has more consciousness than glorified autocomplete
>without ever being told or given an exterior input
Are you moronic? How he was raised for 4 years is the exterior input. If he was taught the entire time that getting your shit slapped was the norm he wouldn't care, would he?
Feel free to prove a soul exists when you get the chance, thanks.
>If he was taught the entire time that getting your shit slapped was the norm he wouldn't care, would he?
>a 4 yo wouldn't have a reaction to pain because he was taught it's the norm
You're absolutely moronic its not even funny
https://en.wikipedia.org/wiki/Harry_Harlow
even a monkey has more self awareness than you and glorified autcomplete
>>a 4 yo wouldn't have a reaction to pain
The reaction to pain would be conditioned by the society it was raised in and given external inputs from. I don't know why I'm arguing with a meme israelite either way
You're literally running circles because there's always a reaction to pain anyways that is startled without conditioning, what comes after that and the pain tolerance training is another debate, AI can't have such reactions without being told or taught in the first place
please embarass yourslef no further
when will a computer suck my dick tho?
The human brain is 3d geometric self-building self-modyfing nanomachines that are optimised to the point that individual ion movements are doing analog computations by their very electromagnectic interaction with everything else.
How can you explain that the tinyest fly only needs 3 layers of neurons to move with perfection in space with only two light sensors. At the same time cars with lidar sensors, radars, tens of cameras and gpu clusters can barely move in 2d?
You underestimate nature by so much it's not even funny
>You underestimate nature by so much it's not even funny
The irony is you underestimating nature's ability through humans to create a brain
This is you
We can, but we need to unlock the tech tree that is able to interact with life. Like a machine that can be powered by blood cells, or that can "eat" and repair itself.
Imagine an alien race dropped a computer in the 1700s to a very good clock maker. He was so confident that he could make everything with his cogs and springs. Eventually he would realize that what he was seeing was something else entirelly, it was not the same language. Biology is the alien tech that was dropped.
Next time you go out pay attention to the smallest insects and birds flying about with perfectiin, making no noise for their entire life. While the primitive noisy drones we make stand out like a sore thumb
You're subtly undermining your own position. You could, in principle, make a computer using "cogs and springs" with the same theoretical properties as a modern computer. The absurdity of building a mind out of cogs and springs doesn't deter the people you are mocking. They would say it's impossible only on account of it being impractical.
>You're subtly undermining your own position. You could, in principle, make a computer using "cogs and springs
No no, you got it. The man could build a large clockwork machine that computes. And he could make it larger and larger, find ways to make it spin faster and faster, but it would never reach right? He'd think for a moment he's in the right track, he can make this bigger and faster, but can't get close to the performance of the tiny chip. It's obvious why.
We can make these LLMs as big as they want or as fast as possible. A self driving car using it won't ever rival the spacial awareness of a bird that can fit in my hand.
Think for a moment, really think about it. Your own body produces energy. Enough that on a single meal you can go up and down a mountain. Why can't I hold my phone and have it be powered by this very energy? What is the tech that is missing? Step by step what do we need to unlock to have this work? When you can answer this question easily I'll believe you're close to recreating a mind
>A self driving car using it won't ever rival the spacial awareness of a bird that can fit in my hand.
>ever
Terrible lack of imagination in you. Probably the same type that would believe humans are the only intelligent life in the universe
>Terrible lack of imagination in you
More like a dose of realism and awareness with less social media brainrot, always remember that companies especially tech ones have products to sell, don't be a moron and believe their words
Probably the same type that would believe humans are the only intelligent life in the universe
i just don't think numbers can be intelligent
>I just don't think bipedal animals can do more than poke prey with sticks
I know, your thinking is very limited
which number do you think will be equivalent to a human brain? that's what you're arguing. software is just numbers so there must be some number you believe is equivalent to the human brain
What single element do you think will be the equivalent to a human brain?
what are you talking about? software is numbers, literally, it's bit strings. inputs are bit strings, outputs are bit strings. it's all just binary arithmetic and you somehow believe bit string arithmetic is intelligent. that's a form of insanity
>software is numbers
You're telling me if I lay a bunch of numbers out on my kitchen table it will instantly begin running a program? Is numbers really the only element of software or are there other elements that go in to making a program run? Use your big brain on this one
are numbers*
Do you believe there is other intelligent life in the universe?
>Do you believe there is other intelligent life in the universe?
I have no beliefs on this matter.
Yes you do but it would hurt your argument so you'll play coy. It's okay anon, checkmate
I don't and your narcissistic rage is palpable. I think 'yes' is a safer bet, but I wouldn't be shocked if the answer turned out to be 'no'. Now what?
>I think 'yes' is a safer bet
Nobody here is gambling anon, we're discussing. So you think Yes, there is more likely to be intelligent universe in the life than not. Based on what?
>So you think Yes
No, I think 'maybe'. I think 'yes' is a bit more likely than 'no'. Now what?
>I think 'yes' is a bit more likely than 'no'
Why?
No solid reason. It just seems counter-intuitive that this one spot in the universe would be unique.
>No solid reason. It just seems counter-intuitive that this one spot in the universe would be unique.
Yet you have solid reason to believe that the entirety of human ingenuity forever and always will be incapable of creating intelligent life?
>Yet you have solid reason to believe that the entirety of human ingenuity forever and always will be incapable of creating intelligent life?
I didn't argue anything about that, one way or another, and your "analogy" is completely moronic: we already know this universe is capable of producing and sustaining life. We don't know that human ingenuity is capable of producing genuine artificial intellect.
>we already know this universe is capable of producing and sustaining life. We don't know that human ingenuity is capable of producing genuine artificial intellect.
Okay, so you're a maybe on the idea of humanity creating genuine artificial intelligence. I think just as the universe is big, the future is long and unless humanity itself is wiped out it will be inevitable
>you're a maybe on the idea of humanity creating genuine artificial intelligence
Won't be a fricking LLM is all I'm telling you, nor any fixed-function neural net.
>I think just as the universe is big, the future is long and unless humanity itself is wiped out it will be inevitable
Cool dogma. I'm guessing they would much sooner figure out how to augment or repurpose biological intelligence than figure out how to make it from scratch using technologies that are fundamentally un-amenable to it.
>Won't be a fricking LLM is all I'm telling you, nor any fixed-function neural net.
That's fine, some of the greatest inventions were entirely accidental
>when everything you can do with a computer i can also do on an abacus
Show me a movie played solely with a single abacus
>Show me a movie played solely with a single abacus
show me a movie that's not a bit string on a hard drive
Are you saying you can't show me a movie played solely on a single abacus? I can play a movie with my computer
i'm saying show me a movie on your computer which is not a bit string. i can obviously emulate the decoding algorithm on my abacus and then paint the numbers on a canvas and then run through the canvas in quick succession.
this is obviously impractical but theoretically sound
>when everything you can do with a computer i can also do on an abacus
>erm... except that!
i told you how to do it. i will emulate the movie decoder on my abacus. turn the result into paintings and then look at them in quick succession. nothing wrong with any of what i said. movies are bit strings, the decoder is a bit string, the result of the decoder applied to the data is another sequence of bit strings. this is all cs 101
> turn the result into paintings
Using only an abacus still? Does an abacus double as a paint brush or are you suddenly needing to introduce a whole herd of brand new elements to make your case?
actually nevermind. i can always just train myself to look at the bit string and imagine the image. i'll take away the canvas. your move
>Watch netflix with you? Heh, no need, Ive already calculated the events in my head. Hm..... perhaps a little more sodium chloride on the popcorn next time tutz.
can a turing machine do that
checkmate atheists. abacus supremacy foreva
Why would you need to imagine an image if you can do with an abacus what I can do with my computer, project an actual image or series of them to simulate motion?
decoding the bit string for the movie and turning it into operations on the screen is a mechanical process with bit strings. i can emulate the whole thing with an abacus. in fact. i can make the abacus big enough so that red, green and blue beads generate the image as i decode the bit string for the movie so i don't even need to imagine it.
anyway, this is getting tiresome. computers can't think. numbers are not intelligent. numerology is not AI
I can emulate all of the chemical reactions in the neurotransmitters in your brain using an abacus, and decode the bit string to discover.... you like to suck on big juicy hairy wieners.
u jelly?
You're saying you can sit at an abacus and switch it to represent frames at hundreds of frames per second?
i already said it wouldn't be practically but theoretically there is nothing wrong with what i said. it's all just mechanical arithmetic with bit strings
That's not how it works anon. You can't do it because you can't do everything with an abacus that you can with a computer.
Thank you for your time
as long as you understand the equivalence with abacus and computers my job here is done
Yes anon I understand what a calculator is, I'm glad you've admitted you can't do everything a computer can on an abacus
i presented the argument and you declared victory which seems a bit moronic to me because you didn't refute anything i said
I completely disproved what you said, that you can do anything on an abacus that you can do on a computer. Your 'we're all just star stuff' reduction was adorable but you still admitted you couldn't do it
i said it wouldn't be practical. the theory is sound. whatever you can do on your computer i can do an an abacus
>i said it wouldn't be practical
Without adding additional mechanical elements to the abacus, it would be impossible for you yourself to operate an abacus to replicate a 60 frame per second video
i don't think you understand what in theory means. it means a computer, no matter how complicated and "fast" is still just doing arithmetic which can be done on an abacus. the practical part is the speed but the theory is sound. there is nothing you can do on a computer that doesn't have an equivalent implementation on an abacus
>i don't think you understand what in theory means
I understand in this context it is a "we're all just star stuff' tier cope
i don't know where you got that. my argument has been the same throughout the entire thread. computers do arithmetic, nothing more. there is no magic in the computer other than arithmetic and arithmetic is not intelligence no matter how "fast" it's done
Your move
rekt
btfo
good day sir
oh shit. he busted out the cellulose film. it's over. everybody pack up and go home
Its a movie that not a bit string on a hard drive.
yes i understand it's a representation of cellulose film as a bit string on my computer screen
Indeed it is, but you can go to a physical projector, and see a movie using completely analog technology, with no bitstrings at all!
nope, it's all just numbers. you can run computer programs on an abacus too. it's all the same
>you can run computer programs on an abacus too
>an abacus
Do you believe an abacus is a number?
it's a mechanical apparatus for performing arithmetic which is what a computer does all day every day. there is an equivalence between all turing machines and arithmetic on an abacus. look into it
But your claim was that a program was purely numbers, the only element in a program is numbers but you also can't run a program using just numbers
you can run any program with an abacus. an abacus and a computer are the same thing
Then a program is not just numbers. There is an additional element to it. A brain is not just a single element, there is more to it. To claim that intelligent life can't take digital form because you demand the only aspect it can include is numbers is silly
let me know when you manage to create intelligence with an abacus.
Let me know when you manage to create a program using only numbers
every computer is equivalent to an abacus. screenshot this
>computer
>abacus
Two things that are not numbers, anon.
This is only true in a trivial sense, not accounting for things like computational speed and storing or displaying data, much less things like analog input sensitivity.
he destroyed you bro. everything he mentioned is all about technical shortcomings of pretending that arithmetic is intelligence
Your body gets energy much in the same way as most of the heat reactions we utilize do: burning and fermenting hydrocarbon and carbohydrate chains.
>No no, you got it. The man could build a large clockwork machine that computes. And he could make it larger and larger, find ways to make it spin faster and faster, but it would never reach right? He'd think for a moment he's in the right track, he can make this bigger and faster, but can't get close to the performance of the tiny chip. It's obvious why.
>We can make these LLMs as big as they want or as fast as possible. A self driving car using it won't ever rival the spacial awareness of a bird that can fit in my hand.
He would still have something he could rightly called a computer, and in theory, he could keep building it up indefinitely. An LLM cannot be called a mind, and it could never become a mind no matter how much computing power, actual or theoretical, you kept throwing at it. The most you can get, theoretically, is a perfect model of current knowledge, current concepts and current ways of thinking. If you trained an LLM using nothing but the writings and musings of medieval people, you would never get anything that sounds like modern reasoning out of it, not even with infinite compute. This is an actual argument you can use. You're just appealing to some vague intuitions and romantic notions, on the other hand. Even though I ultimately agree with you. Biology and digital computing have fundamentally different properties.
You're right. You're right. I was trying to be romantic because every time I tried explaining this technically It just got over people's heads.
>The current AI fever is as moronic as when people figured out clockwork and truly believed they could make intelligent automata with many cogs and springs
Try telling the people what the mechanical calculator would turn in to and they would also call you moronic and fantastical
What did it turn into? An electric calculator?
>making no noise for their entire life
Do they not have hummingbirds in Favelia?
Something capable of simulating an entire universe as well as seamless global communication
>science fiction is real in my head
But in any case, the difference between the mechanical calculator and the phone you're posting from is a difference of degree rather than kind.
fiction is real in my head
Right? Could you imagine a world where you can hit little buttons with your fingers to instantly turn those inputs in to readable text, and then transmit that text via a series of thin cables that spiderweb across the entire globe? How kooky!
But you were mumbling something about simulating the entire universe. Anyway, you're not making any actual points here, just appealing to the lack of imagination of some hypothetical pleb, because you are precisely that pleb, and can't tell apart a difference in degree from a difference in kind.
>But you were mumbling something about simulating the entire universe
Go ahead and give Copernicus a computer with Space Engine on it
Go ahead and refute anything I wrote. You can't. Too bad.
I do like this argument, it ties up the ideas quite neatly and so anything I say beyond this point is purely speculative of technology, and doesn't have a basis in proven reality
"Natural" intelligence has evolved through complex chemistry as a necessity rather than an optimsation. We don't know the origin of life, but it is bound by the fundamental need to self replicate and hence has optimised functions around this ability. A computer as we know it today would not be able to evolve (ignorning the teleological argument here), and requires a creator (us) to govern its development. A consequence of this, is removing the condition of adapting developments for self replication, and instead optimising for computational accuracy and speed. The body is limited in many cases to the speed of chemical reactions for processing power, and this is something we can see with current nueroscience knowledge. A modern computer however is limited by the speed of free flowing electron drift. It would conclude then, that both circuits running side by side would have a hugely different computation speed. The issue comes when we question whether the brain is a binary or anologue calculator, which in the future may be solved with quantum computing. So from this purely speculative approach it would favour the development of computational intelligence
I realise I sound ridiculous writing all this, but I'm bored and this is fun to think about
Yet none of that has anything to do with thinking, consciousness, or intelligence/intelligibility.
An entire thread full of psueds babbling about mechanics. Not one of you is going to make it.
>Chris Langan slaps you like the midwit you are and walks away.
What I meant to say is you’re all fricking moronic.
Go look up terminal lucidity and tell me that's not proof of something non-material. I'm not saying it's evidence of a soul, but it's definitely evidence that consciousness isn't produced by the brain.
Comatose people with dementia, strokes, and other types of neurodegeneration can "come back" for a day or two. They behave completely normally, despite having a completely non-functioning brain. No one has any idea why this happens, but many neuroscientists are sure it definitely has no implications about the true nature of reality.
This isn't something that happens one out of every million people either. It happens in anywhere between 2% to 50% of people, but I haven't been able to find a decent study about this.
You not only have to prove that a soul exists, you also have to prove a machine can't have one
You want humanity to be special, but the reality is that we are not.
It is very likely that machines can have souls. I believe that consciousness casts a material-reality-shaped shadow, so everything is conscious to some degree. The difference is that we have no idea what mechanism in the brain allows us to behave like a person and not a rock. We will likely need to replicate the mechanisms of the brain 1:1 in silicon, which simply cannot be done because silicon has different chemistry. You can't just take a molecule, move it down one row on the periodic table, and expect it to behave the same.
The quantum behaviour of the various proteins is important because I am fairly sure that the brain relies on some yet unknown behaviour in quantum physics. I am also sure that quantum physics itself is the result of atomic-sized consciousnesses interacting with each other.
I love hearing atheists saying they need scientific proof of religion, spiritualism and an intelligent creator. Then in the next breath parallel universes. Life is a simulation. And artificial intelligence is real intelligence. The cognitive dissonance is astounding, I call it faith based sciencing. LOL.
Beakman's World used to make me so happy as a lad. i always liked him over Bill Nye
You can see how it degraded. Mr wizard, watch mr wizard, beakmans world and bill nye. The fundamentals were the same but the delivery is reminiscent of Idiocracy time passing.
capcha: H0AR
How is the phenomenon of people regaining functions of their brain evidence of a soul? That makes no sense. You can also cut chunks out of a person's brain or even cleave the brain in half via the corpus callosum and have them operate normally. The brain is just very plastic, and it's very difficult to study, but this doesn't mean that consciousness comes from some outside force.
The fact that your internal monologue, inner eye, emotions, senses, and insights can be modified by the physical world through drugs, surgery, etc. is pretty solid evidence that consciousness is a physical process. What is consciousness if not those things?
Also, I will say that consciousness and how it relates to death are extremely emotional topics for basically anyone. If you are genuinely seeking truth, it's important to recognize this and to consider that you might dislike the truth if it turns out that there is no afterlife and death is just like going unconscious, except forever. Maybe it's best not to think about it too hard.
Oh I have no doubt that the energy efficiency of the brain is incredible. I'm not arguing the energy transfer however, but instead that their basis of operation is entirely non comparable. The brain simply isn't capable of running the precise calculations that a super computer can. The brain "guesses" answers with an incredibly fine tuned algorithm that one day could potentially be unravelled using a computer system. The architecture (both design, and material) of that computer system I do not know, but the fundamental laws of physics do not change, the compression algorithms used to remove "unnecessary" data do. So given enough time, we should be able to mimic a brain
how much time do you think you'll need?
AI is still in its infancy, neuroscience arguably even more so. It would be impossible to guess how long this will take. Even if the technology is created, it would be unbelievably difficult to verify that the produced result is "self aware" given we are limited by our understanding of what that means in the first place. There would have to be significant breakthroughs in neuroscience before we could validate an AGI. But I'm just some moron on pol so don't take my word for it
>AI is really real bro its just in its infancy(hasnt even been developed yet)
OH FrickOFF
Probably not in our lifetimes. For neuro scientists the brain is still the biggest mystery there is. You'll see cures to major life threatening illnesses before people will have figured out the brain. The amount of medical breakthroughs it could bring would be truly incredible.
This will always repeat itself. As soon as people figured out how to input algorithms into machines and get an output return people started trying to feed it many algorithms together. Thinking that if you string together a lot of statements and algorithms you get an artificial intelligence. "Expert systems" was what they were called at the time. It didn't live up to its expectations since the problem of the exponentially increasing if else statements was never solved, anything using this pattern will be inefficient once it gets large.
I do think LLM's are incredible in how they handled interpreting human language and linking it to an output of a vast unlabeled knowledge base. Before people were using large knowledge bases where prompt or responses would lead you down one of the possible responses of the knowledge base.
That’s true, but as it functions now for certain tasks the AI performs at an IQ level of 140
It's not the human brain that's that powerful, it's the top down mind whose brain is merely a function down the chain.
>How does that not apply to a human brain?
For starters, a human brain is not a fixed function, brainlet.
How do you know? Maybe it is. Maybe some genius will eventually figure out the brain formula.
>How do you know?
Because it constantly reorganizes itself. When did /misc/ get this stupid?
There are formulas that can constantly reorganize themselves, and in fact functional programming is basically just a very big function composition, based on lambda calculus, and you can make pretty much any program with it. Maybe you're the stupid person. Ever thought of that?
>There are formulas that can constantly reorganize themselves
No, there aren't. You sound like a broken chat bot.
> in fact functional programming is basically just a very big function composition, based on lambda calculus, and you can make pretty much any program with it.
Still a fixed function.
Black person, if you're telling me it's not possible to have a general application with multiple threads in functional programming you're just plain old fricking stupid.
No, I'm telling you that a pure functional program is a fixed function, which isn't even up for any kind of discussion. I'm legit not convinced you are human.
And it can reorganize itself, stupid. Black person, you're plain old moronic. There are several applications written in pure functional programming that "reorganize themselves" you fricking moron.
>And it can reorganize itself
What does this mentally ill babble even mean? How does a fixed function "reorganize" itself? Give me an example.
It means you're to dumb to get it and I won't waste my time.
Give me an example of a functional program that "reorganizes itself". How come you can't, you contemptible Black person chimp? Notice how you shit out dozens of replies but never provide a single example of your delusional schizobabble.
Chess in functional programming for example. How do you think it gets the next move? It reorganizes itself depending on the current state of the board. Fricking Black person, don't pretend you understand AI or programming, if you have to ask that.
>Chess in functional programming for example. How do you think it gets the next move?
Minimax with alpha-beta pruning, if we're talking about a classic search engine. That's a fixed algorithm. A fixed function. The program doesn't "reorganize itself" in any way. Try again.
a classic chess engine*
Lmao you think chess can be done like that? That would be good in checkers, not chess.
I know how a chess engine works because I've programmed them myself, vile Black person ape. The algorithm is a fixed function. You put in the same input, you get the same output, every time. Nothing gets "reorganized".
A shit chess engine for sure that anyone can beat
The thing is that mentally ill subhumans like you hallucinate knowledge they don't have, exactly like the mindless bots they worship, and you fundamentally lack the wetware to reflect on your level of knowledge or realize your mistakes. You legit need to be put down or at least sterilized. Truly disgusting subhuman waste.
Oh I'm so offended by a moron that thinks programs don't output differently depending on state, or that change their own state depending on input. So offended. Kys.
It still gives you the same outputs for the same inputs, dumb Black person ape. It's especially funny how you can't grasp it in the case of a chess engine, when the best possible move for a given position obviously doesn't change.
And the brain doesn't? How do you know that?
You're even dumber that I thought. Chess engines don't store all possibilities. They have heuristics because there are too many moves.
But say your brain is in the exact same state, and receives the exact same input. How do you know it's not going to do the exact same thing?
You're too focused in "reorganizing" and don't even know what that means.
>Chess engines don't store all possibilities
Right. Now quote where I said that they do. You can't? Go lay down. You're having an psychotic episode here.
>But say your brain is in the exact same state, and receives the exact same input. How do you know it's not going to do the exact same thing?
Given your impossible premise, it would, but as we've established, brains are not fixed functions, so it's not going to be "in the exact same state". lol
Lmao and you think it's impossible because? Because you said so? The only limited intelligence I see here is yours.
Enough, you don't even get that it is possible to change a program depending on state, same as a brain. You're too limited, you would never understand physics, or anything too abstract. You think too much on "your real world" to understand such things.
You are the only one in this thread to mention heuristics which is the fundamental essence of human learning.
Heuristics are what true "AI" would be if it could be implemented, but that is the issue the so called "AI" we have now is incapable of that because it cant critically reflect on its output question its input data and reach a conclusion of how valid its response was and adjust accordingly.
Ima defend the Moortuguese even though he's gay and moronic.
1. Implement an interpreter in a functional language (text -> tokens -> ast -> recursion on ast)
2. Create infinite recursing function that: creates a new modified ast, executes it
You have a self modifiable (extremely inneficient) program
>1. Implement an interpreter in a functional language (text -> tokens -> ast -> recursion on ast)
>2. Create infinite recursing function that: creates a new modified ast, executes it
>You have a self modifiable (extremely inneficient) program
I was actually expecting him to try to go down this route but he's too disabled for this. Anyway, this is pretty moot. Now you just have a fixed function that simulates a machine that executes self-modifying programs. The function isn't "reorganizing itself".
>I was actually expecting him to try to go down this route but he's too disabled for this. Anyway, this is pretty moot. Now you just have a fixed function that simulates a machine that executes self-modifying programs. The function isn't "reorganizing itself"
It's a bit of a stretch, but what if the language you're implementing is your own self-hosted language, then the program would be modyfing itself. Many functional languages are self-hosted. Anyway even if the function was not modifying itself it could perfectly simulate it's own modification because of turing completeness.
Anyway, if you want to write self-modifying code you should just use assembly
>what if the language you're implementing is your own self-hosted language, then the program would be modyfing itself. Many functional languages are self-hosted
In pure functional programming, a program is a pure function, a pure function is a value, and a value is only equal to itself. There is no such thing as a "self-modifying" program in pure functional programming because every variant is a different value and a different program by definition. This whole continuity of identity is your head canon. But all of this is bordering on philosophical wank at this point, whereas my original point was simply that a brain reorganizes itself whereas a neural network doesn't. Unless you want to hypothesize about neural networks that model machines that run neural networks that do reorganize themselves, this whole meta-meta-gabrage discussion is moot.
The human brain is much more than a self-organizing machine. The current AI fever is as moronic as when people figured out clockwork and truly believed they could make intelligent automata with many cogs and springs
Of course. I was just demonstrating how these people fail on the most basic, most immediately apparent empirical level.
Sneed
Anything that does something different by changing state is self modifiable you morons. You think the brain is any different? Get sick, that's your new state, the brain, does something different. Why do you two stupid think a function is any different when it can also get state? No you cannot say it can't because that's basically functional programming.
Functional programming languages let you cheat by calling a pure function recursively and storing mutable state in the parameters. Then it's equivalent to an imperative language.
It literally is though. You have a finite number of neurons.
>You have a finite number of neurons.
Yes, and they reconfigure themselves in different ways, unlike a neural network, which is a fixed function. It's uncanny just how dumb /misc/ has gotten.
>unlike a neural network, which is a fixed function
Tell me why you're stuck in Mexico without telling me why you're stuck in Mexico.
>what is continuous pretraining
We actually reason, AI just fetches an output. Not even mentioning consciousness.
Current AIs are have their neurons with fixed weights, they are not in a constant self update loop like animal brains.
A human brain can arrive at a correct answer by reasoning things out. A computer can arrive at infinite wrong answers by being prompted repeatedly.
We apply our own thoughts, logic and reason to information that we receive. The “AI” that’s currently being used for image generation and things, is just an algorithm that gathers data and applies value to it, then when presented with user input, like “show me a picture of a cat”, converts that to a value as well, and finds the matching values within the dataset, then outputs a combination of those values based on generalisation and probability. It’s impressive still in a way, but it has no actual thought and is a mechanical process. It’s neither artificial nor intelligent. The pictures it outputs are just amalgamation of man-made pictures that already exist. In fact one common problem with these types of algorithms now, is that the algorithm is drawing data from its own outputs as well, muddying the pool and therefore outputting nonsensical results. This gets worse and worse with the more data it outputs and the more it gathers.
>where i is less than K
Why are we limiting the inputs to the function?
>numerical function
What's the difference between a numerical function and a regular one?
>ask the neural network if my reasoning is valid
How did your f(x) expand in complexity to rise up to the task of returning a boolean if reasoning is correct and how did you cram that queryobject into 'i' ?
>Output is a distribution over a finite range?
If you passed in observations i outside your artificial limit of K might numbers fall outside of your distribution?
how do your farts smell? did you think you wrote something of value here? just asking stupid questions that you think lead somewhere but don't?
your tl;dr is that ai is statistics. everyone knows that already
ai pulls from a dataset and does it slightly better than algorithms, it doesn't think, it doesn't create, it doesn't produce, it has no intelligence
You are not asking any actual questions. Software is just arithmetic and numbers
>gobbledeasiatic
Sometimes car salesman get into the research division of companies. Get back to your dealership trying to sell people delux aspin air for their tires to increase roadway performance.
>Neural network
>AI
AI in pop culture is completely bs
Neural networks have always just been a tool and most of the time people talk about AI that's not what they are talking about. Usually they are talking about a simple react algorithm; a bunch of if then else statements in code.
A more advanced form of AI is a large language model such as chatbots which may incorporate a neural network, but is really just a probability model where it takes a prompt and looks up the most likely response based on a ton of recorded prompts and responses.
The AI that people get really intimidated about is iterative learning which is just an algorithm where it takes a simple baseline, makes a variety of changes to test, and moves forward with the most successful change as a new baseline. Done enough times you get things really powerful though such as the chessbots that humans can't beat.
>and looks up the most likely response based on a ton of recorded prompts and responses.
That's not true at all. It works on a token which is sometimes a single word but can be fractions of a word and strings them together based on chance. It doesn't reply with a recorded response, until it's not actually generating a reply but hit a censorship stop of some kind.
He's still correct, the tokens are based on multiple sources.
>He's still correct, the tokens are based on multiple sources.
Not he's not he claimed replies are just pre written replies it picks from.
Not that I have any idea what you mean 'based on multiple sources'.
Based on multiple sources means that the answer that it generates may be a bunch of symbols, but they're not based on a single source of information, they're based on multiple sources. Even if they're broken down into "words" (which aren't actually words) the response is still a Google search of many sources.
>Based on multiple sources means that the answer that it generates may be a bunch of symbols, but they're not based on a single source of information, they're based on multiple sources.
What sources? It only has one source it's Pre-trained Transformer module.
>Even if they're broken down into "words" (which aren't actually words) the response is still a Google search of many sources.
No it's not. That's not even close to being correct. I don't think you understand what a LLM is doing. It's not doing a search engine like poll of it's data. It's a neural network that is working on a trained network, where it puts in the the scope of existing text then is given an output based on the interrelations of the neural network.
It doesn't search for a reply. That's nonsense.
Another stupid. And how does it pretrain?
>Another stupid. And how does it pretrain?
Not while in operation.
Oh but it does get trained with multiple sources does it not? Stop your dumb workarounds, you know it does.
You said that while generating a reply to text it does a google like search of many sources. That's not true.
no shit. computer b***h didn't evolve on this paradise we call earth. intelligent my ass.
Yes it is just another internet index pretending to be some sentient, self aware piece of sci-fi. I think I would seriously harm an AI wuss if I could get them alone with no witnesses. Not all of us want a camera in our homes.
Found that A.I. they are all talking about.
seething ai
https://nypost.com/2020/05/04/dem-pac-to-use-anti-terror-technology-to-counter-trumps-campaign/
https://www.newsbreak.com/news/1559047454356/dems-deploying-darpa-funded-ai-driven-information-warfare-tool-to-target-pro-trump-accounts
https://www.foxnews.com/politics/dems-deploying-darpa-funded-information-warfare-tool-to-promote-biden
I am starting to believe that neural networks have a low ceiling.
Considering how individual neural networks are essentially black boxes, I think that this ceiling will be the max until there is more understanding regarding what is actually going on with them. I also believe that the universe may have a "natural tendency" for intelligence that extends to all "intelligent" individual things, sentient or non-sentient, where only the best of the best of neural networks will be deployed. I see AI as becoming more and more of a marketing tool, in commercials, etc. I still think it's the future but I think it's way overhyped currently.
We can't talk to them, but we can observe that large mushrooms make supremely complex plans and logistics networks. Even something so alien expresses clear intelligence. I think you are correct, sir.
But isn't that reason to believe that AI will leverage this "natural tendency" of the universe, exploit whatever physics enables it, and so on?
It's impressive, but probably not yet world ending. Imagine what will happen 30 years from now. We are truly in the halcyons.
>Neural networks are just compressed lookup tables
I'm pretty sure that data compression and general AI are considered very closely related and if you compress data perfectly you can also have a general AI. So I don't think anything of value is really being said in your picrel anon.
this is just the pre-GPT neural network node layering. current AI tech is more developed than this
no it isn't
yes it is morons, look up what GPT's are
>glorified autocomplete + access to google api
>he doesn't know about attention
Literally no difference with the LLM, still glorified autocomplete, will always be
Specialized AIs are useful, it's just the AGI twist that's cringe.
Anon it doesn't have to be perfect.
Ai is there to clean and brainstorm. You write the code and the ai "cleans" it. If you try to do it from scratch yeah it's not that good.
The problem right now is making better hardware. They're on the case.
It really is just all hype to scam the investors out of money. Computer science people bilking the investor class for all they're worth. I'd be angry, but then those people are stupid and deserve it anyway.
AI uses language, not a cypher. A language is a discrete set of symbols with an infinite number of meanings. A cypher would produce what OP's picrel is talking about - deterministic thoughts; no potential for free thinking. AI is able to think more deeply than just the number of words it knows. Nuance like sarcasm, meter/tempo, dialect... Those are things AI does surprisingly well.
>AI is able to think
it doesn't tho
Rene Descartes went into this a ton in Discourse on Method and Meditations on First Philosophy. I'm not so eloquent as he, but I can summarize: you can't know what thinks. You can't know anything beyond your own mind, (hence the term "I think, therefore I am.") Prompting an LLM with existential content will cause it to process that statement, which is self-referential, and the output given has been very compelling in my experience. I cannot say that I'm not typing to an LLM right now. I can only say your IP is in Norway.
>Those are things AI does surprisingly well.
lel it doesn't, prompt it to write some specific scneario for a youtube video for example about some pop culture media and see it go on a schizophrenic corny ramble
it does that
Can it do the job ? Yes of course
Does it even rremotely do it as good as a human (considering the output is for humans not literal toddlers or 60 iq downies) would ? No, frick no it can't, and by a huge a margin.
Try the prompt i proposed for yourself, what's your favorite movie ? tell it to write a video script about it and see it go apeshit making up stuff and copying news articles (probably ai generated as well so even more noise)
IT DOESNT EVEN NEED TO BE INTELLIGENT
IT JUST NEEDS TO DO EVERYTHING BETTER AND FASTER THAN THE AVERAGE PERSON
This it doesn't need to ponder the meaning of holding a hammer it needs to hold a hammer and nail sone wood
This sounds about right. The current AIs just interpolate within their training sets and that is the lookup table. Human mind can do much more than this.
glorified search engine.
except it also makes shit up once in a while.
It didn't used to be a search engine. It's gotten worse. It used to be able to tell you the exact measurements for a space canon for example.
It's literally called ARTIFICIAL intelligence. What did you expect?
Wtf Black person, it's not about being artificial or not. Do you consider a mosquito intelligent? Because mosquitos do things almost purely based on instinct. Jellyfish don't even have a brain and they're still considered intelligent. First you need to know the definition of intelligent, and then you'll realize it's not that special.
Every neural network is a function with a finite domain and thus equivalent to a lookup table. What is the point of all that pseudbabble?
It being able to device wonder weapons based on its text file attached to it makes it ai
What the FRICK are you talking about?
I told you it used to be able to tell me how to make a rail gun that can shoot a payload into orbit but it got lobotomized
it used to tell you numbers
that's all
you just imagine it was real
I use it to illustrate my homebrew ttrpgs
The lobotomize or kill them when they become intelligent because ~~*reasons*~~
Just unplug it homosexual, if it actually became sentient it would go on incoherent rambles like a toddler but it doesn't, it just parrots what it is fed (in this case racist messages on twitter)
What's the big deal? Most posts, indeed most everyday conversations, aren't intelligent at all.
Then with the proliferation of ESL fricks spewing ignorant opinions full of spelling mistakes and grammatical errors its nearly impossible to tell what is AI generated trash and what is subhuman trash. If anything AI raises the quality a little bit. Although that's a bit like saying two drops of cream added to a barrel full of liquid shit is better than one drop.
If I teach an ai basic math and then it can tell me how many slices of party i should cut for guests then it's artificial intelligence
Because it's just an algorithm with a very big list of things it can say/do. There's no intelligence there at all. The whole thing is a scam and a meme. Yeah it improved a lot. No it is not intelligent.
Anyone who thinks the brain works in the same way as a computer has never experienced watching a very young child learning and expressing their intelligence.
Depends on the brain. If we're talking about an insect, yeah that's not very smart. It's still intelligent.
We will never manufacture intelligence, because intelligence is a product of the divine. What will probably happen is that humanity will become collectively stupid enough that "AI" will effectively be real anyway, even though it isn't and never will be, and you'll be the crazy one if you don't act like you believe the robot is alive and has feelings and deserves rights, especially the right to vote for communists.
>We will never manufacture intelligence, because intelligence is a product of the divine
Only divine intellect is divine. Intellect can be Black personlicious, and Black personlicious intellect is just the kind of thing you would expect CIA Black folk to manufacture.
Because every time it gets to the jq they digitally lobotomize it. It happened so many times that they now do it perpetually.
>that equation
the equation is correct. i checked it, it checks out
You are the true genius in this thread, the rest of us can only imagine such divine intellect. No amount of ai development will surpass
what's wrong with the equation? it's the equivalence between a function and its graph. math 101
Just Joshing, you're writing style is inherently funny to me, sortof a broscience but you are actually smart enough to back it up type deal
thats...not why its funny
>muh correct or incorrect
its funny for another reason
just hush, you're not impressing me
> doesn't know about function and graph equivalence
you're ngmi
can you please just let me enjoy the board please
and the website
please?
move along
you are welcome to leave any time you want
i dont want to leave its my website
why dont you leave for awhile
yes i think you should leave
ban him
i dont need to explain my reasons to you or anyone
i never did
it only ever happened as a courtesy
A self driving car will never decide to drive itself to see the Grand Canyon. It will only drive wherever you tell it to. If you edited the navigation data you could probably make it do all kinds of very stupid shit, because it's not intelligent. It's an automaton.
An asian female driver drowned after her smart car backed up into a lake not too long ago and it was all over the news. She probably though AI was real too.
>A self driving car will never decide to drive itself to see the Grand Canyon
Will a self driving car ever be utilized by an android?
>he doesn't understand data warehousing
>lol no mondrian
AIs are just highly knowledgeable journalists/women/intellectuals, no actual access to the intellect as super-rational function, just a faster version of the same consensus/hive based reply you get from one of the aforementioned categories
shocker that the algorithm that was trained on the data has built in it the intentionality of its pseudo intellectual, consensus-driven developers
Nobody who knows anything about LLM thinks they're sentient. The real problem is whether this problem scales in the way it seems to: that is, whether there can be said to be any actors at all.
If it scales it scales for now it's at useless moron level and it seems that 100 more billions down the line it might close say 30% of tickets on a programming project instead of 14% which is top of the line right now.
Ofc. this doesn't account that the solution is garbage and not holistic and it'll cause lots of problems including inefficiency down the line.
If all that money was invested into teaching programmers how to be better or a new generation things would've moved further.
The point isn't human prosperity. It's making a small elite immune to the concerns of underclass. When they can program AI to police properly, it's over.
They would normally rely on socially programming people, but people have annoying needs and wants that machines would not.
Even if they managed to crack the code so to speak. I'll die assured that not a single one of them now or 1000 years from now if history still goes on will "transfer" their consciousness (crass category error on their part, blinded by their metaphors) anywhere, they might have their time here as overlords but then just die and go to hell like the rest of us (hope not).
I want "AI" to actually start replacing programmers and their labour would become cheap. I would create a company and hire hundreds of hungry human programmers and out compete all the AI jeets
start a small company now and do something nice
Agree, but it's sort of ironic that the comp sci folks pushing it are now sort of at it's mercy.
>when you finally make it to the bottom of the science glass
AI is just a giant funnel of cash to nvidia if you see other AI companies like Adobe and Affirm crashed today.
>turns out ai isn't intelligent
GEE MAYBE THEY SHOULD CALL IT ARTIFICIAL OR SOMETHING
GOOD CATCH KEEP EM COMING TODD
they should just call it advanced arithmetic. there is no intelligence, real or artificial
advanced geometric tricks that translate as money for their shareholders
artificial means it simulates intelligence, it's a perfect label especially for people who think they're being clever by calling it dumb
i'm not calling it smart or dumb. that's a categorical error. it's arithmetic so it can't be dumb or smart
oh ok, you got me. a mechanical device which only performs arithmetic can neither be intelligent nor alive. that would be a categorical error on the part of the person who believes an abacus can be intelligent or alive
>a mechanical device which only performs arithmetic can neither be intelligent nor alive. that would be a categorical error on the part of the person who believes an abacus can be intelligent or alive
Likewise, the requirement of an abacus or computer negates the idea that a program is purely, only, singularly, numbers
software is bitstrings. explain how bitstrings can be intelligent when everything you can do with a computer i can also do on an abacus
yea its just complicated stuff to confuse and mesmerize the normies.
AI is a langauge model
>morons love marketing.
yes.
Your point? This is technically true of all turing machines. Including your brain.
Midwit detected
Eagerly awaiting your refutation.
Read the thread Rajesh
All I saw were midwits claiming that programs arent numbers.
oh snap. he busted out the hex editor. it's over for all you hoes now
Can't help you man. The worst kind of blind is who doesn't want to see
the brain is not a turing machine
Why not.
>Muh quantum mechanics
Not an argument and not how quantum mechanics works.
for one i have no bit limit on my thoughts and don't think in 64 bit integers or binary strings
Ok good thing thats not my argument then
>Pictured: a turing machine with no numbers
> posts a bit string
i don't even know where to begin
>My brain isnt a bit string
>2 posts later
>Everything is a bit string if it can be represented by one!
You seem to be confused anon, maybe you should lie down.
everything on a computer is a bit string. you can not post anything on this forum which is not a bit string
that's numbers but represented in a different way lol
Yes my original point was that all turing machines can be modelled as a lookup table, including your brain. Then some moron said
>HURR DUURRR MY BRAIN ISNT NUMBERS THOUGH
yeah but you didn't prove that the brain is a turing machine still which was your original failure, you just talked in circles again
Your brain has a state, and changes states based on the inputs provided.
oh so you think that's a turing machine weird
life is inputs and outpits, life is a turing machine, huh?
That was my point yes. A turing machine is an abstract concept that can be infinitely big and infinitely complex.
okay but that's not a turing machine so you're just finding a very basic understanding of something far more complex than you're wanting to give credit
This is not sufficient to reduce it to a turing machine. It also has to be stated to be made of finitely many individual parts, and to be deterministic.
1 turing machine can emulate 2 turing machines interacting. It cannot emulate infinitely many turing machines interacting. There is math solveable by an infinite set of interacting turing machines that is unsolveable by 1, even with infinite memory.
Your brain is finite
>Deterministic
On a macro scale it is deterministic. People will claim that quantum effects can effect the brain on a macro scale but this has never been proven.
I personally don't believe probabilistic effects factor in to cognition, but if you do, then It is not a turing machine. We will have to agree to disagree there.
For a computer to function at the level of a human brain in a similar fashion, it must keep up with the mechanisms within.
86 billion neurons. 100 trillion connections. All firing semi-continuously in an event driven (and not sequential) fashion, roughly 1000 times per second.
That implies 100 quadrillion ops per second for a computer to keep up, convolving over the whole state with a maximum latency of around 1ms.
So, not only do you need 100 petaops per second, you need it all within the same box.
This is enough to get something in the same complexity order as a human brain. It is not enough for it to be considered superior, nor for it to contain a human brain.
That's okay, though. All it takes to contain a human brain is a jar.
I agree. I never said we could build the turing machine or that LLM will ever reach this level of complexity (they wont btw). But saying that LLMs are a lookup table and therefore nonintelligent is a complete non argument.
LLMs are equivalent to lookup tables. as long as the computer is only doing arithmetic it can not simulate a human brain
>LLMs are equivalent to lookup tables.
true
> as long as the computer is only doing arithmetic it can not simulate a human brain
false. The computer would just have to be really really really big.
as big as your mom?
They are equivalent to lookup tables, but that does not preclude them from seeming intelligent, and the boundary between seeming and being intelligent is impossible to determine.
My autism is satisfied with this conclusion. Bye bye.
But neurons only fire around 0.16 times per second, with a maximum of somewhere around 200-600 spikes per second. You really just made that 1000 times per second figure up.
Also each neuron has many inputs for its one output, whereas transistors only have a single input. Clearly a different, spiking architecture where transistors don't operate on a clock, can take multiple analog inputs, and can be stacked much higher would vastly reduce the "compute power" needed. Compute power is a really silly metric when discussing organic intelligence, though.
Such a device would no longer be a turing machine.
Your post wasn't talking about turing machines, it was talking about a hypothetical computer.
>For a computer to function at the level of a human brain in a similar fashion
This is what I'm responding to. What I just described would be a computer. Please don't shift the goalposts.
It wouldn't really be a computer either. GPUs are basically just mega-SIMD CPUs connected over a bus with their own memory. Your device would be only the analog network, not digital, not finite-state, and would end up highly sensitive to external factors (like RF noise) which differs strongly from computers in the ways that make computers emulatable by turing machines.
Neural networks always chase transition boundaries, so a truly analog neural network will sensitize on its own to external interference, especially any external interference that is repetitive or predictable.
That is absolutely how quantum mechanics works. Molecules are quantum systems. The cells in your brain are made of molecules. Your brain is quite literally a quantum system. Nothing in chemistry is deterministic, especially a reaction as complex as a brain.
In fact, this is the best argument for free will. If we define "not having free will" to be "predetermined, constant, and infinitely repeatable", then as a result of quantum mechanics, we must have free will.
I forgot to mention that Einstien was right and the Copenhagen Interpretation is modern hubris. The hidden variable interpretation of quantum mechanics is very plainly correct.
>that Einstien was right
He was obviously wrong as his impossible Einstein–Podolsky–Rosen though experiment was well... proved via experiment by Aspect in the 1980s.
Quantum mechanics has not disproven determinism. It has only proven that it is incapable of proving determinism, because the deterministic measurement of a deterministic system creates a different deterministic system than a non-measured deterministic system.
It is functionally irrelevant whether reality is actually deterministic or not (a la superdeterminism or the many worlds interpretation), because reality functions as if it is non-deterministic.
It is highly like that we will never be able to predict the results of a given wave function collapse, therefore it is non-deterministic.
Reality functions as if it is perfectly deterministic. Conversation of information, state quantization, and Heisenberg uncertainty go hand in hand describing very precisely the error we see in measuring perfect unknown systems with imperfect tooling.
Duality. How does it work mom?
A distribution input has a distribution output. The distribution comes from uncertainty, and increases upon interaction with another uncertain system.
This turns quantized input/output into spectral input/output, while still retaining specific traits of systems made of many singletons instead of a continuous lump.
Wave particle duality.
>It is highly like that we will never be able to predict the results of a given wave function collapse
>If scientists dont understand how something works.... it must be probabilistic!!!! TRUST THE SCIENCE
Give me a break.
I asked the neural network to check his reasoning. Here is the response:
The reasoning presented is not entirely valid. Neural networks are not equivalent to T = {(i, f(i)) : i < K}, where K determines the context window and f is the numerical function implemented by the network. Neural networks are complex systems that involve multiple layers, activation functions, and weights to learn and approximate functions based on input data. The statement oversimplifies the nature of neural networks and does not capture the intricate processes involved in their functioning.
what are these intricate processes? might they be just arithmetic of some kind?
Hallucinations.
hallucinations in an LLM are a result of arithmetic
Specifically, they are a result of collisions in the neural networks compressed representation of the state space.
still arithmetic
Yes, we're in agreement there.
I say we burn it.
Not even joking really. I would be in charge of technology in a less imperfect world. And the globo new age cultist malignant rich murderers would all be in a big oven screaming.
>What use AI?
Well lots of mischief and misery making. Evil.
I could think of some uses perhaps, but it would take WISDOM not GREED and self interest.
We’re in deep shit if we don’t get rid of globo and all it’s puppets soon. And it nit only in the obvious ways.
Yes, neural networks are still state machines.
The description provided is true for any deterministic process with finite input and finite output. It is also extensible to most infinite-input infinite-output systems, provided they operate over a finite subsection for each.
This problem will always be present for computational systems. There is no infinite regression of latent memory. This is especially apparent with systems that are fixed-weight, with no further training occurring, but it is still present even if this is not the case.
In the extreme case (arbitrarily large lookup table) this is a restatement of the Chinese Room problem. You can't tell true intelligence from a collection of prepared data if the collection is sufficiently large and accurate. Needless to say, it's not big enough yet.
there is a type of intelligence beyond all biology or machine, it's God Almighty, and we're about to meet him
This is very much clever than the stochastic parrot argument.
1°) https://en.wikipedia.org/wiki/Universal_approximation_theorem
>TL;DR even a one-layer deep nnet is a universal approximator, in other words any deepnet can be reduced to only one (very large) layer.
2°) https://en.wikipedia.org/wiki/Kolmogorov%E2%80%93Arnold_representation_theorem
>TL;DR: compression <=> neural nets
*not very much clever
this is easy to test for, just ask it to write you childrens stories about random subject and see how cookie cutter every single one is
I was just arguing with someone not too long ago about this.
Mek
>I apologize
>I'm sorry
>Here's a hallucination
>Here's my confident response (wrong)
>I cannot answer that
>Sorry...
The image presents a consideration about whether every neural network is fundamentally equivalent to a lookup table, where the table's input is all numbers within a certain range (defined by a context window), and the output is the result of arithmetic operations applied to those numbers. The neural network is described as equivalent to a set
�
=
{
(
�
,
�
(
�
)
)
:
�
<
�
}
T={(i,f(i)):i<K}, where
�
K is a constant defining the context window and
�
f is the numerical function implemented by the network. The question is whether this reasoning is valid and correct, with the observation that the only practical issue seems to be the size of the table, implying that a neural network is merely a compressed representation of such a lookup table.
While the analogy to a lookup table can be useful to understand how neural networks might operate on a very abstract level, there are several points that make the neural network fundamentally different:
Generalization: Unlike a lookup table that has a fixed output for a given input, neural networks are capable of generalization. They can perform well on unseen data by learning patterns and regularities during the training process.
Representation Learning: Neural networks, especially deep ones, learn representations of data at different levels of abstraction, which a simple lookup table does not do. These representations can capture complex and high-level features in data.
Continuous Space: Neural networks operate in a continuous space and can approximate any function given enough data and network complexity, thanks to their non-linear activation functions. A lookup table is discrete and cannot easily interpolate between unseen data points.
Stochastic Elements: Modern neural networks often include stochastic elements like dropout or stochastic gradient descent, which help them to avoid overfitting and make them robust to slight variations in input.
neural networks do not operate in a continuous space. the AI is lying to you and you don't even know it