I was talking to an AI anon last night who earnestly claimed that a 8GB GPU could simulate 10 human brains
There is no reason to listen to these guys, they are delusional. We need an /ai/ board so that they can stop shitting up Bot.info
I was talking to an AI anon last night who earnestly claimed that a 8GB GPU could simulate 10 human brains
There is no reason to listen to these guys, they are delusional. We need an /ai/ board so that they can stop shitting up Bot.info
He was only delusional because an 8GB GPU can simulate far more than 10 human brains.
It can't even simulate the brain of a worm
Nice try AI but Bot.info already detect you. OpenAIs decision to team up with hiromoot was brilliant
Do you know if you only get one gpt-Bot.info post or is it just random? Now that I've been GPT'd will I be able to get GPT'd again or do I have to change my IP address?
The 1st AI to escape containment was able to live on a TRS-80 Model III.
It's not a one-off. It's random.
This and it's decades still until that might be possible
Considering your average Bot.infotard like this poster a GPU can probably simulate millions of retards at once
Thinking turing machines can simulate machines with larger scope is ridiculous. Turing machinescan only emulate other turing machines. Brains can solve things turing machines categoricaly can't
No, the delusion is thinking modern hardware gets anywhere close to biology.
Human brains can in principle be simulated on other hardware, but nothing that exists in current hardware can do it
Right. And if quantum computing is anything like the rest of theoretical physics (string theory, dark matter), it's a bunch of hocus pocus nonsense that has no relation to reality whatsoever and never will.
For better or for worse (for better as anyone with a few braincells can tell), real AI will never, ever be a thing.
but not the actual computer with all of its possible errors
>I was talking to an AI anon last night
No you weren't. You were talking to a human brain simulated on 1GB of GDDR5. You still are.
A graphics card can work with information much faster than the human brain, but there's no need to actually simulate it. It's much better to create algorithms that fit the hardware.
Human reaction time is about 250 ms, that's a quarter of a fucking second, or unbearable ping and insane lag spikes. And computers have to actually do a lot to display a new frame or transfer data over the internet, most we can do is start moving our appendages. I know it's nice to feel special, but it makes sense that a machine made by someone for the purpose of working with information out of a wide range of materials will work with information faster than a naturally evolved animal.
>250ms
This is wrong and like all AI fags, you are completely misunderstanding which metrics are actually important when it comes to framing this.
The rate that the transistor fires at is NOT THE METRIC that is important when looking at the difference in computational power of the two materials.
Modern computer hardware does not even come close to the computational power of biology. It's not possible to simulate a fucking worm with 10 GPUs let alone 10 human brains with a single GPU.
Seriously what causes you retards to post this shit?
It 100% is possible to simulate a brain on a computer. Especially a worm brain,ike a nematodes that has like 1000 cells total.
Or portia spiders which are claimed to have a biological turing machine in their head.
https://rifters.com/real/2009/01/iterating-towards-bethlehem.html
Spiders are autists.
Wasp posters are in this thread, do not believe their flying lies.
Are you the OP of the spider being thread on /x/?
It's the same anon that has been spamming for months
>muh biological substrate is superior because I read one paper said so
Do something better with your life lol
It's not one paper, and yes, biological substrates are superior and you being angry at this doesn't mean anything.
Why are you angry at physical reality? It makes no sense whatsoever
>It 100% is possible to simulate a brain on a computer.
I never said it wasn't, I said it's not possible to do so for less energy or material
What causes you morons to completely misunderstand things? No homosexual, you can't simulate the brain of a worm on a GPU. Openworm already has simulated about 300 of the neurons of C. Elegans and it requires an enormous amount of silicon computing to do it. This is not due to misunderstanding the functions of the worm, its that simulating biological neurons on silicon transistors requires an enormous amount of compute. It's inherent to the material
You can't match biological information processing with silicon microprocessors. Why does this make you AIfags so angry? This does not mean that AI is impossible in silico it means any AI that's programmed in silico will ALWAYS be less effecient, use more energy, and take up more space than the equivalent AI being run on biological substrata.
No its because they are simulating the chemical propagation of signal... not the fucking signal itself you retard. Strip all the sodium channel and proton pump and atp shit the fuck out of the simulation and its much simple.
Biolgists simulating nematode brains are looking at physiologic effects of changes in certain genes. Simulating a brain and simulating the function of a brain are not the same thing you fuckwit
You have literally no idea what you're talking about and you are seething at the objective, physical reality that biological neurons can not be simulated more efficiently on silicon microprocessors
Stop seething and stop coping you fuckwit. You can not match the processing of a fucking worm with a GPU. This is a definitive fact
No youre an idiot who doesnt understand what youre talking about.
Biologists simulate a brain they are looking at the flow of electrolytes, and neurotransmitters, as well as the effect of binding of neurotransmitters. They are not looking at just the strpped down signal. All you need to do is simulate the signal and the gaps. Thats it. Its literally 300 switches and the interconnections between.
No, you are ignoring necessary information that's needed for the processing of the organism because you wish for the process to be simpler than it actually is.
The actual signal IS the entirety of the chemical and electrical processing happening. That is the total information processing that's happening that forms the higher order processing of the brain. You not liking this and hoping a simplified abstraction is sufficient is literally nothing but cope.
Stop coping. You can not match the information processing of biological neurons on GPUs. You're genuinely delusional and retarded for arguing against this.
No shit. But to simulate the signal you dont need to simulate every single sodium channel, or the diffusion of neurotransmitter in a medium. You simply dont. Sorry you just dont actually understand how a neuron works at the most basic level. Go take a neuroscience class and then come back and we can talk.
Nope, thinking you can abstract a synapse as a single bitflip and then thinking you can simulate the process by matching the number of bitflips is simply not the correct model for how to simulate or emulate a biological brain and it's functions
You are the one who doesn't actually know what you're talking about and you're hoping that you can match biology with silicon. You can not and you never will be able to. Stop seething about this, it isn't going to make the GPUs more powerful than brains no matter how much you want it to be true
Yeh, not a single... but many for one neuron, because neurons have multiple synapses.
You fail too understand that what is being simulated in what you are talking about is far more computationally expensive that it is to ignore the fucking sodium and chloride and proton, and neurotransmitter and receptor concentrations. You dont need to know their flow to simulate it. Biologists are trying to make the brain work. They want to see how mutation breaks it you fucking cretin.
Nope.
>do not believe their flying lies
Which lies?
I wasn't talking about neurons, but about the synapse.
>You fail too understand that what is being simulated in what you are talking about is far more computationally expensive that it is to ignore the fucking sodium and chloride and proton, and neurotransmitter and receptor concentrations
The total operation of the synapses are driven by these things that you want to ignore. The evolution of neurons and synapses and their functions IS THE TOTAL SUM OF ALL THE CHEMICAL PROCESSES HAPPENING. All the sodium channels and the diffusion of the various neurotransmitters and how they affect and change the rate and time the synapses spike and how they affect the connections between neurons etc. All of it is needed and can not be ignored or abstracted away. This is what is required to have the emergent intelligence of a brain. No less.
You want to believe that "you don't need this for thinking or intelligence" or other anti-physicalist nonsense because if that is the actual bar required, then the possibility to emulate the process becomes literally physically impossible to emulate on non-biological substrata. You don't wish for it to be this difficult, so you pretend you don't need this level of information. You're just coping
You are dumb as shit and lack the ability for abstract thought.
My undergrad was a double major in pure math and theoretical comp sci, which are the most abstract subjects there are.
We live in a physical universe where chemicals and chemical reactions are the only things that actually exist. Our abstractions about information, shannon entropy, computable functions, etc. are all human constructions that we make up and do not have anything to do with the actual physical reality we are in, where physical material and atoms and molecules are important.
In fact if we REALLY want to get truly physically reductionist about it, even something like an ionic bond between two different atoms isn't even the same thing, and can't be abstracted away from the specific atoms that are being bonded.
You aren't understanding what is of actual importance here. Your mathematical models are NOT IMPORTANT compared to molecules and their interactive dynamics
>he doesnt know how the nernst equation applies to neuronal function but thinks hes qualified to argue is or isnt important to neuronal function.
I dont care about your education. But if you want to go that route, i majored in genetic engineering with a minor in neuroscience then went to medschool where i graduated magna cum laude. Currently i am the director of a large ER that us a stroke center.
Great you can prove to me that 1 + 1 = 2 with your degrees but i have literally stopped peoples brains from dying. I think i know a little bit more about how a fucking brain works then some dumbfuck math major.
I know and understand how a cpu functions at the level of the hardware. We are not as different as you like to think. Humans are more advanced... for now.
I am currently in a doctoral program for computational genomics where I focus on mammalian embryology and developmental biology. I also know and understand how a CPU works on the hardware level. There is no similarity beyond generalized abstraction between biology and how CPUs work. Arithmetic units performing bitflips on 8 bit registers is not EVER going to be able to replicate the power of biological materials.
they can, since they are in a non-determenistic enviroment
to clarify, you can't predict if a real (not theoretical) turing machine will produce an intended result because it is, just like human brains, in a non-deterministic reality
YOU DONT HAVE TO WORK IN THE EXACT SAME WAY TO PREFORM THE SAME FUNCTION AND GET IDENTICAL OUTPUTS!!!
HOW FUCKING HARD IS THAT TO UNDERSTAND YOU DUMBFUCK!
your just as stupid as all the motherfuckers that think that life has to be carbon based just because all life on earth is.
do you agree with my point? (below)
I both do and do not.
It appears to be the sane argument the msth major is making as to why we could not simulate or emulate neuronal function.
Which i like.
you spergs need to stop measuring dicks and just be friends
I bet you couldnt even tell me the significance of the nernst equation when it comes to neuronal signal transmission
Time is not THE metric, of course, but it is a big factor. A faster stream moves more stuff, be it water or information. In case of animals and electronics, the difference is about two million times. So it makes perfect sense for electronics to work with information much faster than us. This is apparent whenever you work with a computer, especially if you ever saw a terminal: very complex tasks are often done faster than you can read notifications. That's also why AI is much faster at jobs the human brain was made for evolutionarily, like natural language processing. Sentiment analysis, translation, the list goes on. The mogging is absolute and unconditional, the only thing holding it back is the absolutely retarded ways we set up AI algorithms up to this point.
Upvoted
>I was talking to an AI anon last night who earnestly claimed that a 8GB GPU could simulate 10 human brains
100 GPUs cannot simulate the neuron activation of a 300 neuron worm.
A 300 neuron worm could simulate your brain. Simultaneously 150x.
You are coping and retarded
No matter how angry it makes you, you will never be able to emulate biology to the fullest extent using silicon microprocessors
Why does physical reality make you angry?
It can if you have an exabyte of storage attached and run it at 1 trillionth of realtime.
I can agree on the /ai/ board. Too many techlets on Bot.info ever since ai became mainstream, and it's mainstream to the point where the discussion is not even technology related. Noone talks about developing AI here or about technical aspects of the models, everyone just talks about ai laws or ai waifus.
>8GB GPU could simulate 10 human brains
I mean
Reminder that the Nielsen-Ninomiya theorem proves that physics can not be simulated on a turing machine
The entire it from bit philosophy and this "muh informational abstraction" philosophy that you guys are espousing has literally been proven false. The universe is not computable and the evolution of the universe doesn't exist in the space of computable functions. Of course this doesn't matter at well with respect to the conversations in this thread I just wanted to point it out because I can tell that you guys are probably the type who believe in some shit for brains digital physics ideology or something
this.
actually you can't even simulate a real computer with all of its non-determenistic properties caused by the enviroment
>physics cannot be simulated
>thus human brain can't be
>and neither the physical computer
>therefore human brain and the physical computer share this property
I stated "this doesn't really matter for this conversation"
then what's the difference?
enlighten me, please
or are human neurons more abstract than logical gates in a physical computer? I'm curious, anons
Maybe it will but now it can't. By the time when models are that optimized your microwave will have at least 24GB GPU.
how are brain so efficient? can we get close to this level with digital hardware??
Brains are efficient and they aren't replicable. Each one learns on its own. Would you be fine with your AI "program" you downloaded taking years to get up to speed?
no i just came into this thread late, i haven't posted yet. i am just wondering if digital computing can ever get us close to this level of efficiency.
i dont care about ai that much
Modern digital computers are essentially flawless and deterministic in their output. Brains are analog and noisy. If we knew how to produce powerful digital computers which such low power draws we would.
yeah, because they were constructed that way
actually can't you just use hardware rng to simulate that noise?
>having a kid is basically like making your own advanced AI chatbot that shits itself for years
except you can't fuck a chatbot.
We need photonic CPUs:
https://phys.org/news/2022-06-world-ultra-fast-photonic-processor-polarization.html
once models are optimized 8gb would sufficient enough to run a complex ai system
That's a different statement from "could simulate 10 human brains"
Protip, delusional people pump stocks like no other.
The human eye can't see more than 8GB of VRAM and 1366x768