Training neural networks to emulate human behavior is analogous to distillation, a process by which a neural network can be trained to emulate another neural network. The optimal way to predict the next token a human will say is to have a near perfect understanding of humans and the universe.
Not without proving that consciousness is a computation. Nothing in computer science, biology, neuroscience, or philosophy indicates that you're assertion is true. In fact, Goedel wrote several proves which cast serious doubt over the possibility of consciousness being computational
If it's not a computation, then what, dumb Black person? Fricking magic? Everything in the universe can be reduced to the computation between quantum bits of information.
Quantum Mechanics is Indeterministic.
Random chance is not "computation" - and yes to the layman "luck" and "magick" are indistinguishable.
God controls the RNG. Terry Davis was right.
quantum mechanics are not indeterministic they just rely on the quantum field which can't be measured due to the heisenberg uncertainty, it is the same as throwing a coin it only seems random because we don't know all the variables
>Everything in the universe can be reduced to the computation between quantum bits of information.
The thing is, we don't really know how the universe works. Or more precisely, we only know what we know. There's still undiscovered physics and who knows what else.
>Everything in the universe can be reduced to the computation between quantum bits of information.
The thing is, we don't really know how the universe works. Or more precisely, we only know what we know. There's still undiscovered physics and who knows what else.
probably, but we don't know how gravity can be reduced to computation "between quantum bits of information", for example, we don't know how gravity would affect a superposition of particles, or how gravity can be quantized, that's the "theory of everything" that physicists are trying to find.
The rule book would not really be a single book, and you would need to keep track of millions of variables, if you wanted to actually simulate a person capable of carrying on a conversation while executing "dumb" algorithmic instructions to do so. The "Chinese Room" only appears like it can "prove" that language is not enough to show intelligence by vastly oversimplifying the problem; indeed nobody has ever actually managed to *construct* a purely algorithmic language model that is capable of holding a conversation. LLMs are capable of this but their workings are inscrutable for the most part.
The argument has nothing to do with the exposing the simplicity of what an AI is actually doing but it poses the question of what knowledge actually is. It distinguishes between the action a human takes and the actual intuitive understanding the human has of the action itself. In this case even if an AI did have the capability of holding a full conversation, can we really say that this functionality is equal to human consciousness? The human (or AI in this case) in the room has not shown the ability to actually understand the language, but only the ability to carry it out.
>can we really say that this functionality is equal to human consciousness?
right, intelligence does not imply consciousness
i always thought chinese room was a stupid thought experiment since the brain is already split up into functional modules (there's probably a better word for this)
the BG knows nothing about chinese, but it takes control from the cortex in the form of muscle memory and writes the characters reliably
>the brain is already split up into functional modules
If you cut a person's brain in half it won't really affect their live. It's insane watch this video:
In this case I would say that the person may not know Chinese, but the room does. I think that originally, the sheer possibility of something that wasn't a human "knowing" anything was simply overlooked or dismissed offhand.
>I think that originally, the sheer possibility of something that wasn't a human "knowing" anything was simply overlooked or dismissed offhand.
Because the argument is mainly in response to the turing test.
I bet you don't even know what a dendrite is do you?
moron
i bet you think dendrites are only input
Training neural networks to emulate human behavior is analogous to distillation, a process by which a neural network can be trained to emulate another neural network. The optimal way to predict the next token a human will say is to have a near perfect understanding of humans and the universe.
they think it is literally people, in no small part thanks to all the bullshit sensationalist news and moronations like GPT "talking" through a animatronic torso like they had on some talk shows
that's not how """AI""" works. ML is all about statistics and weights
your post actually shows normal programs. if you look at a diasseembled program, you'll see a bunch of comparisons, integer/FP operations, calls and jumps.
>statistics and weights
statistics are only used for learning, i.e. descent to closest local minima of the network function with respect to the weights
the weights are part of the neurons and the bias acts as an activation threshold which is the continuous version of the McCulloch-Pitts units which are just logic gates
the only difference to real brains is that a single one of our neurons can actually compute the xor function while artificial neurons cant
No it's only you, you're the messiah, the light in the darkness, truley you are the protaginist of reality with your ability to semantically split fricking hairs over a definition you don't understand, thank you OP I want to suck your wiener and tell you how your parents were totally wrong for saying you've wasted your life obsessing over trivialities.
I'm confused by these AI models.. Will they be developed into machines that can truly think fairly soon, or will this period of rapid progress eventually stall well before we reach such a stage?
Short answer, not. Long answer: general AI is simply not possible. It's so impossible that the problem isn't even defined. This is perhaps the most uncontroversial answer in CS. It's only grifters who promote agi as a possibility.
by OPs logic, anything involving modern computers is a bunch of if and elses, and therefore can't be artificial intelligence. he jumps to conclusions without any support for his argument.
by OPs logic, anything involving modern computers is a bunch of if and elses, and therefore can't be artificial intelligence. he jumps to conclusions without any support for his argument.
it's basically the bell curve meme: people who know nothing think it's AI, people who know a little bit of compsci think it's just a bunch of if/else statements, and people who know a lot think it's AI
normal people have been preprogrammed by mass media (like with everything else) to believe AI is the next generation of humanity that will surpass us and drive us to extinction when it inevitably becomes rogue due to [insert movie plot here].
I think mass media has actually been programming us to believe it'll become rogue, like how it's been programming us to think nuclear power is dangerous, when really we've now fricked up the climate by shutting down nuclear power stations so we can burn coal instead. The same thing will happen to AI, which will be pointlessly delayed for years due to stupid Skynet fears, and in the meantime millions of people will die
97746528
I still believe the convoluted way SAO and Cyberpunk (there's better examples) tried to make A.I. is so far the only likely cases and actual human-like A.I. will be ever produced, we can't make animals like Pigs from zero, you need a base made from something already existing. Why would human-like A.I. be different?, if anything the problem is that a human A.I., unless we give it some sort of way to feel things in-computer will likely go crazy from total sensory deprivation kind of quickly, anxiety fricks people because they feel like they physically can't do something, now imagine waking up and being unironically NOT PHYSICAL.
I'm even wary of the possibility that one odd A.I. will be like “nah, I'm fine” and just deceive us till it can actually, have some degree of power or physicality through tools or something else like the Johnny ending or the Alice bot. Or worse.
wow a whole lot of people have very strong opinions in here about how the brain works. I don't know about anyone else but I don't know anything about how the brain works. not a single thing!
Why do you insist that the human genetic code is sacred or taboo? It is a chemical process and nothing more. For that matter we are chemical processes and nothing more. If you deny yourself a useful tool simply because it reminds you uncomfortably of your mortality, you have uselessly and pointlessly crippled yourself.
this is stupid
it's more like a line drawn on a graph
You can literally say the same thing about our own brains
n-no but we have a s-s-soul
>le soul
kek
Not really. Not really at all.
Yes really.
I bet you don't even know what a dendrite is do you?
moron
it's a person who fricks trees
Training neural networks to emulate human behavior is analogous to distillation, a process by which a neural network can be trained to emulate another neural network. The optimal way to predict the next token a human will say is to have a near perfect understanding of humans and the universe.
Not really. No
moron
I can say that about your brain fricking bot
Not without proving that consciousness is a computation. Nothing in computer science, biology, neuroscience, or philosophy indicates that you're assertion is true. In fact, Goedel wrote several proves which cast serious doubt over the possibility of consciousness being computational
>Not without proving that consciousness is a computation.
Roger Penrose thinks it isn't and he's a smart guy.
If it's not a computation, then what, dumb Black person? Fricking magic? Everything in the universe can be reduced to the computation between quantum bits of information.
Quantum Mechanics is Indeterministic.
Random chance is not "computation" - and yes to the layman "luck" and "magick" are indistinguishable.
God controls the RNG. Terry Davis was right.
quantum mechanics are the result of self balancing, its not rng
quantum mechanics are not indeterministic they just rely on the quantum field which can't be measured due to the heisenberg uncertainty, it is the same as throwing a coin it only seems random because we don't know all the variables
>only seems random because we don't know all the variables
Literally disapproved:
https://en.wikipedia.org/wiki/Bell%27s_theorem
since you're so confident please describe to me the physical process through which the human brain produces qualia. make sure to be specific!
>Everything in the universe can be reduced to the computation between quantum bits of information.
The thing is, we don't really know how the universe works. Or more precisely, we only know what we know. There's still undiscovered physics and who knows what else.
probably, but we don't know how gravity can be reduced to computation "between quantum bits of information", for example, we don't know how gravity would affect a superposition of particles, or how gravity can be quantized, that's the "theory of everything" that physicists are trying to find.
retroactively refuted by the chinese room argument
pic related
The rule book would not really be a single book, and you would need to keep track of millions of variables, if you wanted to actually simulate a person capable of carrying on a conversation while executing "dumb" algorithmic instructions to do so. The "Chinese Room" only appears like it can "prove" that language is not enough to show intelligence by vastly oversimplifying the problem; indeed nobody has ever actually managed to *construct* a purely algorithmic language model that is capable of holding a conversation. LLMs are capable of this but their workings are inscrutable for the most part.
The argument has nothing to do with the exposing the simplicity of what an AI is actually doing but it poses the question of what knowledge actually is. It distinguishes between the action a human takes and the actual intuitive understanding the human has of the action itself. In this case even if an AI did have the capability of holding a full conversation, can we really say that this functionality is equal to human consciousness? The human (or AI in this case) in the room has not shown the ability to actually understand the language, but only the ability to carry it out.
>can we really say that this functionality is equal to human consciousness?
right, intelligence does not imply consciousness
i always thought chinese room was a stupid thought experiment since the brain is already split up into functional modules (there's probably a better word for this)
the BG knows nothing about chinese, but it takes control from the cortex in the form of muscle memory and writes the characters reliably
You're begging the question that consciousness is reducible to brain matter. Explain mathematical intuition and knowledge - qualia.
>the brain is already split up into functional modules
If you cut a person's brain in half it won't really affect their live. It's insane watch this video:
We still have no idea how brains work.
In this case I would say that the person may not know Chinese, but the room does. I think that originally, the sheer possibility of something that wasn't a human "knowing" anything was simply overlooked or dismissed offhand.
>I think that originally, the sheer possibility of something that wasn't a human "knowing" anything was simply overlooked or dismissed offhand.
Because the argument is mainly in response to the turing test.
>multiple companies manufacture literal chinese rooms you can talk to
>"n-nobody has e-ever ackshually..."
Chinese room only applies to the hardware, not the rule-set itself.
what dya mean
Meaning that the algorithm running in the chinese room really DOES know chinese even if the people/whatever inside doesn't.
wrong
correct
i bet you think dendrites are only input
pseud
correct
disinformation
truth
That was Boolean's idea yeah
they think it is literally people, in no small part thanks to all the bullshit sensationalist news and moronations like GPT "talking" through a animatronic torso like they had on some talk shows
this is literally how people work though
only weve evolved for thousands of years for doing exactly this so we are pretty good at it
Midwit reductionist thread
>hurr durr it's all just binary, voltages on a wire, AI isn't real
>it's all just binary
not for long
Neuromorphic chips will return to computing in Superior Analogue soon enough
Based. Analog computers are the future.
This video sums everything up
that's not how """AI""" works. ML is all about statistics and weights
your post actually shows normal programs. if you look at a diasseembled program, you'll see a bunch of comparisons, integer/FP operations, calls and jumps.
>statistics and weights
statistics are only used for learning, i.e. descent to closest local minima of the network function with respect to the weights
the weights are part of the neurons and the bias acts as an activation threshold which is the continuous version of the McCulloch-Pitts units which are just logic gates
the only difference to real brains is that a single one of our neurons can actually compute the xor function while artificial neurons cant
The great Tardigrade disagrees
homosexual
No it's only you, you're the messiah, the light in the darkness, truley you are the protaginist of reality with your ability to semantically split fricking hairs over a definition you don't understand, thank you OP I want to suck your wiener and tell you how your parents were totally wrong for saying you've wasted your life obsessing over trivialities.
I'm confused by these AI models.. Will they be developed into machines that can truly think fairly soon, or will this period of rapid progress eventually stall well before we reach such a stage?
Short answer, not. Long answer: general AI is simply not possible. It's so impossible that the problem isn't even defined. This is perhaps the most uncontroversial answer in CS. It's only grifters who promote agi as a possibility.
there are no branch instructions in neural networks
Technically no. But they really aren't that much more complicated conceptually. OP's point still stands
by OPs logic, anything involving modern computers is a bunch of if and elses, and therefore can't be artificial intelligence. he jumps to conclusions without any support for his argument.
Fricking moronic Black person
So are you
>So are you
what the frick does that even mean?
it's basically the bell curve meme: people who know nothing think it's AI, people who know a little bit of compsci think it's just a bunch of if/else statements, and people who know a lot think it's AI
actual Black person logic
normal people have been preprogrammed by mass media (like with everything else) to believe AI is the next generation of humanity that will surpass us and drive us to extinction when it inevitably becomes rogue due to [insert movie plot here].
I think mass media has actually been programming us to believe it'll become rogue, like how it's been programming us to think nuclear power is dangerous, when really we've now fricked up the climate by shutting down nuclear power stations so we can burn coal instead. The same thing will happen to AI, which will be pointlessly delayed for years due to stupid Skynet fears, and in the meantime millions of people will die
Dunning-Kruger: the thread
It's more like a matrix multiplying a matrix.
97746528
I still believe the convoluted way SAO and Cyberpunk (there's better examples) tried to make A.I. is so far the only likely cases and actual human-like A.I. will be ever produced, we can't make animals like Pigs from zero, you need a base made from something already existing. Why would human-like A.I. be different?, if anything the problem is that a human A.I., unless we give it some sort of way to feel things in-computer will likely go crazy from total sensory deprivation kind of quickly, anxiety fricks people because they feel like they physically can't do something, now imagine waking up and being unironically NOT PHYSICAL.
I'm even wary of the possibility that one odd A.I. will be like “nah, I'm fine” and just deceive us till it can actually, have some degree of power or physicality through tools or something else like the Johnny ending or the Alice bot. Or worse.
wow a whole lot of people have very strong opinions in here about how the brain works. I don't know about anyone else but I don't know anything about how the brain works. not a single thing!
Why do you insist that the human genetic code is sacred or taboo? It is a chemical process and nothing more. For that matter we are chemical processes and nothing more. If you deny yourself a useful tool simply because it reminds you uncomfortably of your mortality, you have uselessly and pointlessly crippled yourself.
define intelligence