A neuron, whether in biological organisms or artificial constructs, functions as a fundamental information-processing unit. In both contexts, it operates by transmitting signals through connections to other units, enabling the flow of information within a complex network. This basic unit plays a pivotal role in decision-making and information exchange within its system, contributing to the overall functionality of the larger structure.
>just throw more "neurons" at the problem. Seemed to work for us
No, it didn't. The brain isn't just a pile of neurons. It consists of specialized structures. All the cool LLMs and diffusion models you use also aren't just straight ANNs but have somewhat elaborate architectures.
you don't need to understand human intelligence, you just need to make an architecture that doesn't have the pitfalls of the previous architecture, rinse, repeat.
eventually you will run out pitfalls and it will look nothing like human intelligence.
False equivalence, flight was understood to be moving through the air since forever. We still don't even really know what "thinking" is. There is no way to prove that anyone besides you exists, but you can pretty easily prove that birds are in the air and they need wings, even if you can't easily prove the physical principles behind it.
>There is no way to prove that anyone besides you exists
that's just masturbation.
you aren't smart just because you propose endless amounts of unprovable bullshit.
if you live in reality like the rest of us you would accept the notion of duck typing
>There is no way to prove that anyone besides you exists
yeah there should be, theoretically. but you're not actually interested in that are you? you kinda like the spot, you'll actually fight against any way it could be proven, quite actively, to defend your cozy spot. innit?
Okay you fricking gorilla Black folk I misspoke, I meant there's no way to prove other people exist NOW, and I support looking into it. I'm just saying that designing intelligence is hard if we don't know what intelligence is. We knew what flight is even if we didn't get the details and so we could reverse engineer it. We do have an idea of what intelligence is but it's not as simple as brid go up in air
>There is no way to prove that anyone besides you exists >but you can pretty easily prove that birds are in the air and they need wings
These two are contradictory, either our measurements of the world are just arbitrary to the observer(only you exist) or they're reality(you're an individual among many). We can measure if beings have an understanding of self and ability for creative problem solving.
It also misses the point completely, I was alluding to the fact that humans found a different way to fly than birds, even though some thought that was impossible. We will probably find a different way to consciousness than biology did.
>I was alluding to the fact that humans found a different way to fly than birds
But what I'm saying is that we know that planes fly and birds fly by simple looking at them, while we don't know that intelligences are "intelligent" in the same way that we are by simply looking at them, and we currently have no way to do that. We can only reverse engineer a crude idea of intelligence, which could very well be leading us to real intelligence, but we have no way to verify that like we do with planes
You just setting up artificial barriers, humans have never had to actually understand anything properly to use it. It's all just models, and most engineering is done with grossly simplified models since they're actually mathematically solvable. For what we're doing in machine learning, the current measure for AGI is if the system can learn and solve any cognitive task better than a human. It doesn't matter if it's "actually" conscious, whatever that even means, if it can generalize across problems and solve them. We can't even agree on a definition of consciousness since it's an idea of philosophy, but we can measure if the machine beats you in math, humor, dating, engineering, learning, etc.
You're excruciatingly low-IQ. I'm pretty sure his point is that you can't know if you're dealing with actual intelligence or just an illusion of it if you keep going by vague subjective metrics, so until you have existing examples of intelligence figured out, you can't know if you're even going in the right direction with your attempts. For instance, the LLM fans are obviously going in the wrong direction and they either cannot or do not want to fathom it. They will probably spend decades deluding themselves that their statistical model is intelligent as they throw more and more data at it, training it to be better and better at fooling them.
>flight was understood to be moving through the air since forever.
why the frick did people even attempt to make flapping contractions then?
if it was so obvious, they would have realized birds flap their wings like 9000 times per millisecond and that trying to force a human to do the same, no matter how complicated the cope contraption, was fricking pointless
understanding something does not equal dominating something
>Flight will not be achieved until we understand bird wings
Flight was achieved when the underlying physical principles were understood. Principles that are many orders of magnitude simpler and easier to experiment with.
Not implying intelligence is as easy as flight, just that it was thought we had to understand bird wing mechanics to do it.
We picked a very different architecture after iterating over several, where some tried copying birds, how is that much different in principle from building an ass load of different function estimators until one of them starts giving us solutions to problems.
You're excruciatingly low-IQ. I'm pretty sure his point is that you can't know if you're dealing with actual intelligence or just an illusion of it if you keep going by vague subjective metrics, so until you have existing examples of intelligence figured out, you can't know if you're even going in the right direction with your attempts. For instance, the LLM fans are obviously going in the wrong direction and they either cannot or do not want to fathom it. They will probably spend decades deluding themselves that their statistical model is intelligent as they throw more and more data at it, training it to be better and better at fooling them.
>I'm pretty sure his point is that you can't know if you're dealing with actual intelligence or just an illusion of it
I know, but it's irrelevant. Does it matter if the black box is actually intelligent when it designs novel semiconductor layouts or drugs? No, not at all, you still arrive at the same result.
How the frick is it vague or subjective to test if a system can beat the best human experts at various tasks? If it is better, it wins, the same way humans are tested to figure out who's the best in a certain field. I think the world is all ears if you have a better way of testing superior intelligence than doing cognitive tasks better than others.
LLMs and Diffusion models in their current form aren't AGI, but if you compare them to the absolute dog shit models we had 10 years ago it should be clear we're going in the direction of more capability.
>Not implying intelligence is as easy as flight, just that it was thought we had to understand bird wing mechanics to do it.
What does it actually have to do with OP's point, beyond matching the same linguistic template? Nothing. It's crazy how automatonistic your whole crew is. I can see how you arrivded at this total nonsequitur: you started form a standardized corporate zinger ("they said the Wright brothers couldn't fly, luddite!!!"), saw that the situation doesn't quite fit this preprogrammed template and modified it ever so slightly into something superficially more related.
>I know, but it's irrelevant.
It's relevant and I've proven it in the very post you are replying to. Stop shitting out corporate zingers and try again.
Must be great to be a simplistic automaton, but my point and I assume this is also what OP was getting at) still stands: you can't know how far along you are in the process of building an airplane if you can never tell objectively if you've gotten a single inch off the ground.
>How the frick is it vague or subjective to test if a system can beat the best human experts at various tasks? If it is better, it wins
This is a test of automation viability, not a test of a system's intelligence. If you want to talk about automation, just talk about automation.
I fail to see why there would be a meaningful separation of intelligence and task solving, humans solve tasks with our intelligence, if a single system can solve any task better than any human, it must have a more general and precise world model than humans. Capacity to solve problems is generally how intelligence is defined, maybe you define it differently and if so, please clarify.
>if a single system can solve any task better than any human
And how are you going to build this magical sci-fi system can that solve ANY task better than ANY human if you have no idea what the fundamental requirements for such a system are and whether or not you're on the right track to implementing them?
4 months ago
Anonymous
ummm sweaty? just two more weeks, two thousand more layers, two trillion more parameters, two quadrillion more terrabytes of your data, two more decades of compute time, two quintillion more terrawatts of power... two more weeks and matrix multiplication will become a viable solution to every problem a human has ever faced and will ever face
I think you mean consciousness since we understand intelligence pretty well even with the field being handicapped and scared of admitting biological realities
Plausible
just throw more "neurons" at the problem. Seemed to work for us
Define neuron. No I will not take "node in a directed graph with an assigned weight vector" as an answer
A neuron, whether in biological organisms or artificial constructs, functions as a fundamental information-processing unit. In both contexts, it operates by transmitting signals through connections to other units, enabling the flow of information within a complex network. This basic unit plays a pivotal role in decision-making and information exchange within its system, contributing to the overall functionality of the larger structure.
>just throw more "neurons" at the problem. Seemed to work for us
No, it didn't. The brain isn't just a pile of neurons. It consists of specialized structures. All the cool LLMs and diffusion models you use also aren't just straight ANNs but have somewhat elaborate architectures.
Decision Trees > Neural Networks
>this kills the AI Pseud
How can AI trannies ever recover after reading this post? How many will make the decision to hang themselves from a tree?
we'll use AI to understand it
This. It may be a cheat but we could get around it.
AI was already achieved. Are you talking about AGI? maybe but we've achieved things we haven't fully understand before.
Like english
Indeed
#1 anime of this season, followed by android girl in mars and ishura
>#1 anime of this season
based
>followed by android girl in mars and ishura
cringe
Rough fricking the stupid mole rapist girl
Why would we need to understand human intelligence when we could just get AI to understand it for us?
Utena!
OP you need to address your AGP with therapy before you try to solve the problem of AI.
human intelligence is a myth
much noise itt, this is the only message with a serious philosophical underpinning.
>my IQ is 80
you don't need to understand human intelligence, you just need to make an architecture that doesn't have the pitfalls of the previous architecture, rinse, repeat.
eventually you will run out pitfalls and it will look nothing like human intelligence.
Didn't believe this would get it's own anime, I'm impressed.
there have been worse
You should try and achieve intelligence since AI is extremely common and has been for over 2 decades.
No and remember:
Racism, sexism, makeup, and plastic surgery is living proof the human form is ugly.
You could only prove me right.
Delusional pajeet
who fricking cares
not a single person in that picture looks attractive, photoshop and surgery only made them a different kind of unattractive
That's true but it took them from overtly disgusting to look at to weird-looking but tolerable.
none of them are tolerable to look at tobetbh
You human form is ugly. Don't speak for real humans thanks.
>ESL
https://arxiv.org/abs/2106.05181
It doesn't exist.
We are all just saying what we think other people want to hear.
We were language models all along.
>what we think other people want to hear
frick you
I came
Flight will not be achieved until we understand bird wings
I mean you kinda peeked didn't you?
False equivalence, flight was understood to be moving through the air since forever. We still don't even really know what "thinking" is. There is no way to prove that anyone besides you exists, but you can pretty easily prove that birds are in the air and they need wings, even if you can't easily prove the physical principles behind it.
>There is no way to prove that anyone besides you exists
that's just masturbation.
you aren't smart just because you propose endless amounts of unprovable bullshit.
if you live in reality like the rest of us you would accept the notion of duck typing
>There is no way to prove that anyone besides you exists
yeah there should be, theoretically. but you're not actually interested in that are you? you kinda like the spot, you'll actually fight against any way it could be proven, quite actively, to defend your cozy spot. innit?
Okay you fricking gorilla Black folk I misspoke, I meant there's no way to prove other people exist NOW, and I support looking into it. I'm just saying that designing intelligence is hard if we don't know what intelligence is. We knew what flight is even if we didn't get the details and so we could reverse engineer it. We do have an idea of what intelligence is but it's not as simple as brid go up in air
>There is no way to prove that anyone besides you exists
>but you can pretty easily prove that birds are in the air and they need wings
These two are contradictory, either our measurements of the world are just arbitrary to the observer(only you exist) or they're reality(you're an individual among many). We can measure if beings have an understanding of self and ability for creative problem solving.
It also misses the point completely, I was alluding to the fact that humans found a different way to fly than birds, even though some thought that was impossible. We will probably find a different way to consciousness than biology did.
>I was alluding to the fact that humans found a different way to fly than birds
But what I'm saying is that we know that planes fly and birds fly by simple looking at them, while we don't know that intelligences are "intelligent" in the same way that we are by simply looking at them, and we currently have no way to do that. We can only reverse engineer a crude idea of intelligence, which could very well be leading us to real intelligence, but we have no way to verify that like we do with planes
You just setting up artificial barriers, humans have never had to actually understand anything properly to use it. It's all just models, and most engineering is done with grossly simplified models since they're actually mathematically solvable. For what we're doing in machine learning, the current measure for AGI is if the system can learn and solve any cognitive task better than a human. It doesn't matter if it's "actually" conscious, whatever that even means, if it can generalize across problems and solve them. We can't even agree on a definition of consciousness since it's an idea of philosophy, but we can measure if the machine beats you in math, humor, dating, engineering, learning, etc.
You're excruciatingly low-IQ. I'm pretty sure his point is that you can't know if you're dealing with actual intelligence or just an illusion of it if you keep going by vague subjective metrics, so until you have existing examples of intelligence figured out, you can't know if you're even going in the right direction with your attempts. For instance, the LLM fans are obviously going in the wrong direction and they either cannot or do not want to fathom it. They will probably spend decades deluding themselves that their statistical model is intelligent as they throw more and more data at it, training it to be better and better at fooling them.
>flight was understood to be moving through the air since forever.
why the frick did people even attempt to make flapping contractions then?
if it was so obvious, they would have realized birds flap their wings like 9000 times per millisecond and that trying to force a human to do the same, no matter how complicated the cope contraption, was fricking pointless
understanding something does not equal dominating something
If his point is a bird, then your head is your head.
>Flight will not be achieved until we understand bird wings
Flight was achieved when the underlying physical principles were understood. Principles that are many orders of magnitude simpler and easier to experiment with.
Not implying intelligence is as easy as flight, just that it was thought we had to understand bird wing mechanics to do it.
We picked a very different architecture after iterating over several, where some tried copying birds, how is that much different in principle from building an ass load of different function estimators until one of them starts giving us solutions to problems.
>I'm pretty sure his point is that you can't know if you're dealing with actual intelligence or just an illusion of it
I know, but it's irrelevant. Does it matter if the black box is actually intelligent when it designs novel semiconductor layouts or drugs? No, not at all, you still arrive at the same result.
How the frick is it vague or subjective to test if a system can beat the best human experts at various tasks? If it is better, it wins, the same way humans are tested to figure out who's the best in a certain field. I think the world is all ears if you have a better way of testing superior intelligence than doing cognitive tasks better than others.
LLMs and Diffusion models in their current form aren't AGI, but if you compare them to the absolute dog shit models we had 10 years ago it should be clear we're going in the direction of more capability.
>Not implying intelligence is as easy as flight, just that it was thought we had to understand bird wing mechanics to do it.
What does it actually have to do with OP's point, beyond matching the same linguistic template? Nothing. It's crazy how automatonistic your whole crew is. I can see how you arrivded at this total nonsequitur: you started form a standardized corporate zinger ("they said the Wright brothers couldn't fly, luddite!!!"), saw that the situation doesn't quite fit this preprogrammed template and modified it ever so slightly into something superficially more related.
>I know, but it's irrelevant.
It's relevant and I've proven it in the very post you are replying to. Stop shitting out corporate zingers and try again.
Must be great to have the whole world and everyone in it figured out.
Must be great to be a simplistic automaton, but my point and I assume this is also what OP was getting at) still stands: you can't know how far along you are in the process of building an airplane if you can never tell objectively if you've gotten a single inch off the ground.
>How the frick is it vague or subjective to test if a system can beat the best human experts at various tasks? If it is better, it wins
This is a test of automation viability, not a test of a system's intelligence. If you want to talk about automation, just talk about automation.
I fail to see why there would be a meaningful separation of intelligence and task solving, humans solve tasks with our intelligence, if a single system can solve any task better than any human, it must have a more general and precise world model than humans. Capacity to solve problems is generally how intelligence is defined, maybe you define it differently and if so, please clarify.
>if a single system can solve any task better than any human
And how are you going to build this magical sci-fi system can that solve ANY task better than ANY human if you have no idea what the fundamental requirements for such a system are and whether or not you're on the right track to implementing them?
ummm sweaty? just two more weeks, two thousand more layers, two trillion more parameters, two quadrillion more terrabytes of your data, two more decades of compute time, two quintillion more terrawatts of power... two more weeks and matrix multiplication will become a viable solution to every problem a human has ever faced and will ever face
She's so dumpy and a loser. God I want to be whipped by a dumpy nerdy sadistic girl so bad.
I want to cum in her moronic hair
fake it till you make it
use the AI to figure intelligence?
I think you mean consciousness since we understand intelligence pretty well even with the field being handicapped and scared of admitting biological realities
>AI will not be achieved until we understand human intelligence.
Humans are not intelligent.
Rust is not a good programming language and children should not be exposed to transgender propaganda.
Humans are stupid, self-destructive, rodent hybrids of an even higher lifeform.
Frick em. Frick you.
Education facilities already understand that quite exceptionally well.