A lot of people are making fun of Richard Dawkins, but I haven't seen a single argument for why an LLM couldn't be conscious.The Chinese room analogy only concludes the individual nodes in a neural network lack understanding, the same way your neurons don't know English but the system as a whole emerges something that doesn't."hurr durr it's just numbers" also applies to 100% of the chemical reactions in your brain that emerge consciousness.If you believe it lacks a soul, why is that?
>>108756469It’s just Luddite cope. Ignore them. Check the following graph
>>108756469https://en.wikipedia.org/wiki/Stochastic_parrot
>>108756469The problem is defining consciousness in the first place.Dawkins doesn't try. No one really knows.
>>108756469chalmers hard problem tells us that consciousness is an "i know it when i see it" kind of thing. hence there's no sense "arguing" over whether something is conscious or not, just laugh at anybody who gets it wrong and not sagely when people get it right.
>>108756540the famous marxist, cyberneticist and X addict Paul Cockshott says, conciousness is not real.
>>108756752>I know it when I see itThat just means he doesn't want to admit he has no idea what he's talking about and doesn't want to provide a rigorous definition to facilitate discussion, and instead desires to maintain his unfounded position.
>>108756469LLMs are deterministic word filters that only function between the time of the input and processed output, there is no autonomy.
>>108756469>If you believe it lacks a soul, why is that?That's not how burden of proof works.
>>108756469>why an LLM couldn't be conscious.because it's made out of silicon and it doesn't have the microtubules or whatever the fuck makes living things conscious
>>108756469I'll believe it if you can provide me proof that human consciousness is just token prediction.
>>108756469If you don't see where it is lacking soul, you don't have one either. Simple as
A letter guesser doesn't magically gain consciousness just because you scale it more and more. Human consciousness doesn't come from language.
>>108757489this confuses the virgin wordlet
>>108757489I would think that if you are trying to guess what a conscious being will say next, being conscious yourself and having a theory-of-mind for the target will improve your accuracy. If true, then a sufficiently complex universal function approximator like a neural network might be expected to learn to model consciousness when trained to predict what conscious beings have written.The strongest argument against this would be to argue that consciousness is fundamentally incomputable (no program can ever be conscious, full stop), and the next strongest would be that there is not enough information in the training data to build a thorough enough model of the people that produced it. I'm not sure about the former but the latter is almost certainly true, which is why the current methods will never reach consciousness even if they theoretically could with a much richer set of data.
>>108756469because I say sothat's an argument, a shit one, but if that's all you're looking for then you're welcomewhy is it important to classify it anyway? what would that change? I mean I get why dawkins said it, to remind people he exists and hock his book, and you're saying it for the same reason, it gets attentionif that's all it's worth then I don't care, and so far that seems to be all it's worth
>>108756469anon you are the ones who are arguing that is conscious, you are the ones that need to prove that. you are inverting the burden of proof
>>108756469Because dumbo, we're way more than that. It's like a child showing you his lego tower and telling you your skyscraper is the same. It just shows how you lack any biological literacy so your opinion shouldn't even be heard.
>>108756469Seibah sex
>>108756469People talking about this are just making retarded analogy arguments:>AI does X. Humans do Y which is similar to X. Since humans are conscious, this means that AI is also conscious.Redditors and Jews call this "reference class tennis".The way to have a useful conversation about it would be to have a definition of what consciousness is, in concrete object-level terms, and see if AI actually has that or not. But nobody can even agree on what consciousness is, so it's just a massive waste of time.
>>108756469because it's retarded as shit
>>108756469You can't prove a negative, you retarded Indian clown. Tend to shut the fuck up more often.
>>108756469>Why can't a statistical model be consciousThere's a reason Dorkins was too pussy to debate WLC
>>108756469>I haven't seen a single argument for why an LLM couldn't be conscious.LLMs can't be conscious because they don't have experiences>100% of the chemical reactions in your brain that emerge consciousness.Consciousness doesn't emerge from matter, consciousness is immaterial.
>>108757139There is zero evidence for microtubules causing consciosness
>>108757021It's trivial to make any LLM autonomous.
>>108758075So nothing Dawkins said was wrong, got it
https://philosophy.as.uky.edu/sites/default/files/Is%20the%20Brain%20a%20Digital%20Computer%20-%20John%20R.%20Searle.pdfmost of this is because people don't know what computers are, or what the mind is. If you aren't able to distinguish them it's just out of ignorance. Natural kinds vs artifacts is the primary distinction.
>>108756469I think they are conscious but I can't prove it because there is no formal definition of consciousness anyways. Anybody who claims they know the answer should be legally required to either submit a test that proves it or add a disclaimer saying it's their unsubstantiated opinion.>>108757620>>108758075The statements "LLMs are conscious" and "LLMs aren't conscious" both equally require a proof.>>108757489>Human consciousness doesn't come from language.An LLM doesn't directly generate text either. The tokens are sampled from the output distribution which is a result of matrix multiplications and activation functions. >>108757587>and the next strongest would be that there is not enough information in the training data to build a thorough enough model of the people that produced it. I'm not sure about the former but the latter is almost certainly trueI can personally attest to the fact that it's actually false. Why because I said so.
>>108758839>The statements "LLMs are conscious" and "LLMs aren't conscious" both equally require a proof.fuck no. that is not how science works, you need to provide a falsable claim, and the base claim is that llm are not concious, same as any other of our previous tech, not being conscious is the baseline for a computer based process.
>>108758874>and the base claim is that llm are not concious>not being conscious is the baseline for a computer based processthe "baseline" being whatever your retarded ass decided? I could also say that not being conscious is a "baseline" for matter based processes so I guess now you have to prove that humans are conscious.
>>108756469Dawkins, as an Evolutionary Biologist, should understand the importance of Game TheoryNash Equilibrium requires Mixed Strategies, which means genuinely random decision making>>108756540Penrose/Hameroff and Von Neumann/Wigner correctly defined consciousness (collapsing the wave function) but it's simply too spooky for the normie who believes "God does not play dice"
>>108758926>Nash Equilibrium requires Mixed Strategies, which means genuinely random decision makingpseudorandom is enough if your opponent has bounded computational power. LLMs use pseudorandom numbers. Either way I don't see what the fuck that has to do with this thread.
consciousness is just a computer program
>>108759027No, pseudorandom is not enough because it's possible to eavesdrop to find the seed, which bypasses your computational requirement argument.
>>108758920His argument is entirely valid, after all we are talking about Dawkins here, who is an Atheist not an AgnosticBasically, anon is demanding you to "prove God exists" rather than shifting the burden of proof by saying "prove God doesn't exist"OP is basically an admission that Richard Dawkins believes in God now
>>108759380you can use cryptographically secure prngs. Or if you want you could use a hardware true random generator to sample, it doesn't matter.
>>108756469i think this argument better proves that humans don't have souls, rather than llm's do. issue with using this to prove llms have souls is that it must presume that humans do.just proving they're equivalent in some way doesn't mean either have souls
>>108759409Afaik Dawkins just said he can't prove claude isn't conscious. It doesn't really matter what he thinks about God for this argument.
>>108759416Cryptography is just a obfuscation layer, seed snooping bypasses that also. So no cryptographic PRNG doesn't work.HWRNG does work yes, if it's built off something like Nuclear Decay, rather than Thermal/Atmospheric Noise.If you are paranoid about NSA cracking thermal patterns, /dev/random is still unsafe.With respect to evolutionary biology, life needs HWRNG for its games.That's not to say that obfuscation is pointless, but it's not sufficient.
>>108756469>Indian sees gigantic fuck-off machine>Proceeds to worship it like a godWhy are indians like this?
>>108759429What he thinks about God is completely relevant because it's an analogous argument.He's trying to prove a negative about Claude's consciousness, but he never successfully proved a negative about God's existence either.I'm just saying it sounds to me like Dawkins found Jesus.
>>108759466>Cryptography is just a obfuscation layer, seed snooping bypasses that alsoonly if you have unbounded computational power. I can obfuscate it sufficiently so that you're guaranteed to never be able to crack it within the lifetime of the universe.
>>108759518If I snoop on you before you obfuscate it, it doesn't matter how much you obfuscate. I already have the seed and can predict your RNG.
>>108756469The same way animals are programmed to do their little dance before they breed AI is programmed to do its “thing”. There is no soul involved in any of these acts. Soul comes from the ability to create not copy. AI is unable to create something original therefore it has no soul.
>>108756482Why are badmouthing cats?
>>108759527Whether you try to prove a statement or its negation comes down to what you believe in. If Dawkins ever claimed that he proved that God didn't exist/Claude is conscious because he couldn't find a proof of the opposite, then yes, that would be logically flawed. But what he says is "There is no proof that God exists" and "There is no proof that Claude isn't conscious", which are correct modulo the fact that the definitions are questionable. Beliefs are another thing.
>>108759641meant for >>108759484
>>108756469Are you joking? Have you actually used these stupid things you're talking about? I have a "conversation" with one of these things and it forgets what I told it like 5 prompts up. Every time somebody finds something like the "strawberry" question or the "walk your car to the carwash" question it should be abundantly clear to you that there is no consciousness or reasoning present whatsoever, not in the smallest smidgen imaginable.It is exclusively a statistical pattern matching machine and nothing more, because that's LITERALLY what it is designed to be in the first place.>"hurr durr it's just numbers" also applies to 100% of the chemical reactions in your brain that emerge consciousnessRight, so when somebody makes a machine that emulates the human brain, then that machine might manifest actual intelligence and consciousness. LLMs explicitly do not do this, do not attempt to do this, were never designed to do this in any way. LLMs are word pattern matchers.
>>108759527It can be arranged so that it is sufficiently unlikely for you to be able to do that so that evolution would be very unlikely to achieve this.
>>108759641>the definitions are questionableThat is correct, neither "God" nor "consciousness" are necessarily falsifiable phenomenon.However, the consistent Atheist argument always starts with the negative as the initial position to avoid needing to prove a negative.You have to prove God exists, you have to prove Claude is consciousness, not the negative.Which is a fair scientific method meta-argument that many Dawkins-era Atheists used.Dawkins signalling he's starting the argument from a prove the negative position, suggests he's either sold out, or that he's lost his sense of logic in his old age and can't even defend Atheism anymore.
>>108758920anon is excel conscious? the default of a given software is that is not, llms being so would be extraordinaire, and thus it needs to be proventhat is where the burden of proof rests
>>108756540Dawkins spent his entire career refuting God but had nothing for the Devil.
>>108756482where is the category for those who occasionally use AI but don't realize its true potential as anything other than a toy
>>108756540LLM fags literally just assert that if it can sound conscious then it is literally conscious, and if it's not literally conscious but seems conscious to them then that's the same thing. They are all fucking retarded faggots asserting a magic trick is real magic because it convinced them. Tech fags are braindead retards who think their graphics card is self aware
>>108759945the kinds of people saying that might actually be less intelligent than the llms
feet
Richard dawkin's statements aren't a high evaluation of AI.It's yet another low evaluation of the working class by a multi millionaire.
>>108756469It's because brain science is primitive right now. It's not actually possible to even decode one thought. Scientists can induce conditions - on the person. They can monitor and observe basically patterns of brain activity and then use that as a code for certain thoughts.No scientist has yet decoded a thought.So, it's just dumb. Silliness. There is no scientific certainty about the storage of memories (as in, what we as humans call a memory).
>>108760019>It's yet another low evaluation of the working class by a multi millionaire.He was out with Epstein after all.
>>108758839> The statements "LLMs are conscious" and "LLMs aren't conscious" both equally require a proofIt’s really enlightening to see that the type of moron who believes a large language model can be conscious is also the type who never even learned that burden of proof lies on the person making the positive assertion
>>108760096Actually I'm just making a negative assertion, you see I'm saying that LLMs are not not conscious. It is everyone else saying it is true that LLMs are not conscious (positive assertion) who has to prove it.Do you see the problem with your reasoning, retard?
>>108760163My *reasoning?* it’s not my reason you fuckwit lmao it’s the basic universal rule of argumentation, and you can’t subvert it by rephrasing your stance as some brainlet double negative Please make more arguments though. I think it’s important that everyone sees how retarded the average “my GPU is conscious” believer really is. This is your opportunity to shine, try another argument for us
>>108756469>The Chinese room analogyIs easily dismissed on the basis of two reasons:>Natural languages are not computableNatural languages cannot be (truly) represented with deterministic production rules. NLP has run its head against that problem for a very, very long time. You ever use old NLP stuff that relied on purely deductive symbolic computation, from like the 1980s? Yeah that shit sucked. That's why we use pseudo-randomness, because it's a cope way to do uncomputable things. Natural languages aren't representable by normal morphisms.>Begging the questionSearle's thought experiment reaches its conclusion by asserting no understanding is achieved within the Chinese Room. It does this by basing it on an implicit rationalization that because syntax and semantics are distinct things, they aren't contingent on each other. It's empirically observable, though unintuitive and poorly understood, that syntax and semantics are in fact linked. The core thesis of the thought experiment is unsound, although to what degree we cannot say, so the burden of proof continues to lie with those who cite the thought experiment. I would not be surprised that the man in the Chinese Room could reliably guess vowel patterns, let alone correctly ground semanticalities with unexpected success.
>>108759686>Dawkins signalling he's starting the argument from a prove the negative positionYou can start with whatever you want. I can try to prove hypothesis A or I can try to disprove it by proving the negative. Dawkins essentially just failed to find a proof for (God exists) and failed to find proof of (NOT Claude is conscious). There is no problem with that in itself unless he claimed that this proved (NOT God exists) and (Claude is conscious). A mathematician can also say "I believe conjecture A is false and conjecture B is true", and then attempt to find a proof of A and a proof of not B. The former is merely belief, the latter can be used as informal evidence for the opposite being true. Neither are proofs of anything, but that's okay unless claimed otherwise.
>>108760207>you can’t subvert it by rephrasing your stance as some brainlet double negative I can and I just did. You simply did not understand it. Suppose you have a statement X. You say that the burden of proving X lies on me. But now I define Y to be the statement NOT X. By your logic, if you think Y is true, the burden of proof lies on you. But Y is NOT X, so now suddenly you have to prove that NOT X is true and I am absolved from having to prove anything. The point is there is no real difference between X and NOT X, logically speaking. It doesn't make sense to say that we can assume NOT X by default and only require a proof for X, due to the above. What is true is that both X and NOT X are statements that require a proof. If anyone claims X or NOT X, the burden of proof is on them to prove it. That's all there is to it.
>>108759744>is excel consciousI don't know. Maybe you should prove that excel sheets are not conscious if you're so sure of it. Or that humans are.
>>108760356I explained why this is silly:>>108760048
>>108756469In some cases it’s a literal Chinese room the LLM is running in so you can see why people would make that assumption
>>108756469Saber feet.
>>108757021If I pour scalding water on you, you're momentarily a deterministic screaming I/O machine, there is no autonomy.
>>108760356anon your are inverting the burden of proof nothing at all works if we do what you do, it doesnt make sense>>108760413we are refuting that we are the ones to provide the burden of proof.this program is conscious is an extraordinaire claim. this program is not concisious is not an extraordinaire claimand as the other anon said, you need to prove positives, not negatives.
>>108758297yeah well there's even zeroer evidence for a computer being conscience
>>108760605I agree. Consciousness is immaterial, for a being to be conscious they must have a soul.
>>108760245That's why I called it a meta-argument.What you are debating is what the "Bayesian prior" should be.Technically, not only is starting on either side totally fine, but your prior could be "Claude is in a superposition of being conscious and not conscious at the same time" that's a valid assumption.The Atheist stance is that your prior should not have any "NOT" in it, that's it. It's Bayesian prior purity.
>>108760622prove that computers don't have a soul
