The most prestigious law school admissions discussion board in the world.
|
 |
 
|
Reminder: Searle's Chinese Room is a TTT argument
Poast new message in this thread
Date: July 12th, 2014 5:06 PM Author: jade free-loading station toaster http://plato.stanford.edu/entries/chinese-room/
"Imagine a native English speaker who knows no Chinese locked in a room full of boxes of Chinese symbols (a data base) together with a book of instructions for manipulating the symbols (the program). Imagine that people outside the room send in other Chinese symbols which, unknown to the person in the room, are questions in Chinese (the input). And imagine that by following the instructions in the program the man in the room is able to pass out Chinese symbols which are correct answers to the questions (the output). The program enables the person in the room to pass the Turing Test for understanding Chinese but he does not understand a word of Chinese."
Edit: this has been used to discredit the possibility of an "understanding" AI by Searle himself as well as several others.
Do prestigious "intellectuals" actually buy this crap? When someone asks you a question, and you respond, the language centers of your brain know a bunch of rules for interpreting and responding to your native language...the instructions/exact response comes from other areas of your brain (emotional, rational) coming up with an appropriate response and sending it back to the language centers to parse it and spit it out.
In this thought experiment, the guy in the room is basically serving the same purpose as the language centers of a human brain. He's just interpreting and responding to data based on rules and responses he has been given (in the brain example, other areas of your brain telling the language section what to tell your mouth to say, basically).
No one would argue that the Temporal Lobe of the brain by itself "understands" anything.
|
 |
Date: July 12th, 2014 5:12 PM Author: Awkward violent double fault hall The elucidation of these issues is the point of the thought experiment. (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#25918047) |
|
Date: July 12th, 2014 5:14 PM Author: jade free-loading station toaster What do you mean? The point Searle tried to make with this is "Okay sure you could have AI that passes the Turing Test but it can never actually 'understand things' the way humans do" (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#25918056) |
|
Date: July 12th, 2014 5:16 PM Author: Awkward violent double fault hall tbh I don't want to get stuck defending Searle because he literally said that biological substrate was necessary for intelligence. With zero justification. (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#25918066) |
|
Date: July 12th, 2014 5:17 PM Author: jade free-loading station toaster yeah he's just a biomo, it seems.
that's like a pre-Wright brothers bro saying physiological substrate is necessary for flight. |
|
Date: July 12th, 2014 5:19 PM Author: Awkward violent double fault hall Agreed. (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#25918085) |
|
Date: July 27th, 2014 5:25 AM Author: offensive honey-headed hospital BUT THEN WHO WAS KNOWER (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008671) |
Date: July 12th, 2014 5:11 PM Author: jade free-loading station toaster It's like saying "A microprocessor doesn't understand anything, therefore AI that understands stuff is impossible" (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#25918041) |
|
Date: July 12th, 2014 5:19 PM Author: Pontificating low-t principal's office You're 100% wrong, his argument is quite decisive. His point is that computers that operate using systems of synatic rules do not understand semantic content. They can follow the rules that are fed into them, but they do not understand what those rules means. Maybe a traditional computer could have a "conversation" in chinese if the synatic rules it were using were complicated enough to allow it to act like a Chinese person, but the fact that it does not understand the things that it is saying means it is not intelligent. (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#25918087) |
|
Date: July 12th, 2014 5:24 PM Author: insecure menage Is this to say that his arguments don't apply to artificial neural networks that don't operate by carrying out syntactic rules?
I've heard Searlmos argue that Searle did nothing short of demonstrating that only a biological brain could ever achieve consciousness. |
|
Date: July 12th, 2014 5:25 PM Author: Pontificating low-t principal's office His argument only applies to traditional computers, yes. (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#25918125) |
|
Date: July 12th, 2014 5:28 PM Author: 180 sticky national security agency heaven ANNs carry out syntactic rules. what are you talking about? (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#25918134) |
|
Date: July 12th, 2014 5:29 PM Author: jade free-loading station toaster "His point is that computers that operate using systems of synatic rules do not understand semantic content. They can follow the rules that are fed into them, but they do not understand what those rules means."
It's pretty obvious what he is saying, but the argument was proposed to refute the possibility of AI that can "understand" what it's doing when it passes the Turing test. And it's used by AI-naysayers, including him, to show that you can't have "intelligent" or "understanding" AI. Which I'm saying is 100% wrong.
Searle's "Chinese Room" would only represent one part of the AI infrastructure, the language interpreter. By that logic, the human brain is not "intelligent" or "understanding" because your Temporal Lobe "doesn't know" what it's doing...it's just applying rules you've learned. |
|
Date: July 12th, 2014 5:31 PM Author: Pontificating low-t principal's office I agree that it doesn't prove AI is impossible. Anyone that says it does is making a fairly large mistake.
"Searle's "Chinese Room" would only represent one part of the AI infrastructure, the language interpreter. By that measure, the human brain is not "intelligent" or "understanding" because your Temporal Lobe "doesn't know" what it's doing...it's just applying rules you've learned. "
Humans do understand semantic content, our consciousness isn't built merely on applying syntactic rules. It's hard to say much more than that since the nature of human consciousness isn't well understood by either philosophers or scientists. |
|
Date: July 12th, 2014 5:34 PM Author: jade free-loading station toaster "Humans do understand semantic content, our consciousness isn't built merely on applying syntactic rules."
Right. But the individual part(s) of our brain that would be performing the same function at the interpreter in the Chinese Room are equally as non-understanding. So the argument is kind of defeating a Strawman of what developed AI will likely be (if it's possible...which it seems like). |
|
Date: July 12th, 2014 5:36 PM Author: Pontificating low-t principal's office The problem with your argument is that you're making all sorts of assumptions about how the human brain works and how it is linked to consciousness. If we consider what it is to follow syntactic rule it is clear that it involves NO semantic content. A machine with no consciousness at all can be programmed to follow such rules. Something else is going on in our brains or our minds, but we do not know what that thing is (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#25918172) |
|
Date: July 12th, 2014 5:45 PM Author: jade free-loading station toaster "The problem with your argument is that you're making all sorts of assumptions about how the human brain works and how it is linked to consciousness."
We know from fMRI/PET scans of people with and without brain damage to specific brain regions that certain regions (i.e. language processing centers) are very specific and not involved in much other than processing language rules.
"Something else is going on in our brains or our minds, but we do not know what that thing is "
This is true, although it would surprise me if it were anything more than a lot of highly-interconnected parts, most of which are relatively non-understanding individually.
We know from sending electrical impulses to different brain regions that we can "shut off" or "stimulate" certain functions that affect consciousness. We just don't know the details.
Given that if you "shut off" certain areas of the brain, a person will become unconscious, it doesn't seem likely that there are any parts of consciousness that don't result from biology. But it's possible. |
|
Date: July 12th, 2014 5:39 PM Author: Wine theater stage legend dogs can read human faces, and without intellectual comprehension of the analytic content behind those expressions, respond in (empirically derived from trial and error) appropriate ways (conducive to getting treats).
are dogs working within a chinese room framework? |
|
Date: July 26th, 2014 11:15 PM Author: jade free-loading station toaster "are dogs working within a chinese room framework?"
explain. anything that processes, interprets and responds to information using a set of rules could fit within a "framework" of the Chinese room. but whether it can be concluded that that its conscious or not is different. |
|
Date: July 26th, 2014 11:13 PM Author: jade free-loading station toaster (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26007473) |
|
Date: July 27th, 2014 1:29 AM Author: dashing crimson lay can't stand searle, what a fucking boomer striver - actually goes thru the trouble to retake bachelor's during grad studies to wipe off his shitstain B.A. at wiscousin (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008061) |
|
Date: July 27th, 2014 5:06 AM Author: Hairless ruddy institution turdskin pretty common for rhodes/marshall scholars iirc (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008634) |
|
Date: July 27th, 2014 4:54 AM Author: jade free-loading station toaster would like a certain "chill" smart aspie poasteur's opinion on this (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008616) |
|
Date: July 27th, 2014 4:55 AM Author: Hairless ruddy institution turdskin i don't really understand your response. are you saying there "understanding" somewhere in the chinese room? if so, where? (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008618) |
|
Date: July 27th, 2014 4:57 AM Author: jade free-loading station toaster no. i'm saying that yes, Searle's original conclusion of the Chinese Room argument is tcr, but his use of it to discredit the possibility of a certain "understanding" type of AI is unwarranted. Because of false analogy. (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008623) |
|
Date: July 27th, 2014 5:03 AM Author: Hairless ruddy institution turdskin i guess i'm only familiar with the original conclusion, which isn't aimed at the physical possibility of any sort of AI but at a set of propositions he called "strong AI" (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008630) |
|
Date: July 27th, 2014 5:07 AM Author: jade free-loading station toaster so I should clarify here. I really meant that using Searle's argument to discredit a certain type of "understanding" AI (which Searle himself and several others have done) is not credited. Not necessarily that the initial argument itself is uncredited (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008636) |
|
Date: July 27th, 2014 5:11 AM Author: Hairless ruddy institution turdskin that doesn't seem like what your OP is doing tbh (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008643) |
|
Date: July 27th, 2014 5:14 AM Author: jade free-loading station toaster yeah why do you think I clarified?
because the argument is commonly used against AI and that's how I discovered it. I forgot to poast that in the OP though. |
|
Date: July 27th, 2014 5:15 AM Author: Hairless ruddy institution turdskin the extension of the argument against AI is fairly simple ("add another set of rules. well, we've already decided rules can't get you understanding. so you're still screwed") but would apply equally to building a human from the ground up (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008656) |
|
Date: July 27th, 2014 5:20 AM Author: jade free-loading station toaster right, yet humans are "understanding" as per the definition.
so its clear that ..... theres something more. or more likely understanding is an emergent property of a complex system (see above my analogy about the human temporal lobe) |
|
Date: July 27th, 2014 5:21 AM Author: Hairless ruddy institution turdskin or there's something special about the human biological substrate, as searle believes.
"emergent properties" are mostly hand-waving imho |
|
Date: July 27th, 2014 5:24 AM Author: jade free-loading station toaster "or there's something special about the human biological substrate, as searle believes"
yeah there's no evidence of this bro. and no reason to believe it.
unless you're a penrose fag who thinks microtubules create human consciousness |
|
Date: July 27th, 2014 5:26 AM Author: Hairless ruddy institution turdskin not sure what evidence emergence has but i haven't thought about the issue in a few years. at a certain point before these matters are fully explored the evidence will just be people's intuitive reactions which'll differ wildly (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008673) |
|
Date: July 27th, 2014 5:29 AM Author: offensive honey-headed hospital seems right. fwiw i prefer the hand waves of emergence to the hand waves of bio substrate (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008679) |
|
Date: July 27th, 2014 5:29 AM Author: jade free-loading station toaster "not sure what evidence emergence has but i haven't thought about the issue in a few years."
there really isn't any. and anything that talks about conscious experience is inherently doomed to failure.
we were talking about this earlier this evening in some other thread but...its not so important that AI is "understanding" in a fully-sentient human-like conscious sense (which is impossible to prove and shaky to begin with) but rather that it can do .....higher-level thinking that is equal to or greater than are most powerful minds. as SAPIENPIGS |
|
Date: July 27th, 2014 5:30 AM Author: Hairless ruddy institution turdskin "important"?
searle's argument is about semantic understanding and, arguably, through that about consciousness. not about what's important |
|
Date: July 27th, 2014 5:35 AM Author: jade free-loading station toaster "searle's argument is about semantic understanding and, arguably, through that about consciousness. not about what's important"
no, we aren't talking about searle's argument. we were at first, now we aren't. we are talking about consciousness in general.
what im saying is we already concluded that its bullshit and impossible to say anything about "consciousness" and by extension a certain chill human-like "understanding". so its more important that we focus on what we CAN do. |
|
Date: July 27th, 2014 5:36 AM Author: Hairless ruddy institution turdskin how did we conclude that it's impossible to say anything about consciousness. i wasn't at that meeting (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008695) |
|
Date: July 27th, 2014 5:39 AM Author: jade free-loading station toaster its pretty simple. its a qualitative state only you can experience.
anything else is inferred.
edit: its a subjective state you cant qualify |
|
Date: July 27th, 2014 5:40 AM Author: Hairless ruddy institution turdskin so it's qualitative, huh? *takes notes* (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008703) |
|
Date: July 27th, 2014 5:41 AM Author: jade free-loading station toaster are you dumb or something? (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008704) |
|
Date: July 27th, 2014 5:42 AM Author: Hairless ruddy institution turdskin everyone seems to think i am tonight, maybe they're on to something
my point was that "qualitative" seems to be "something we can say about consciousness", which is what you said was impossible |
|
Date: July 27th, 2014 5:45 AM Author: jade free-loading station toaster look im really drunk but ill attempt to engage you further
what i meant was subjective |
|
Date: July 27th, 2014 5:28 AM Author: offensive honey-headed hospital compared to magical substrate? (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008677) |
|
Date: July 27th, 2014 5:29 AM Author: Hairless ruddy institution turdskin more or less equally so as far as i can see (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008678) |
|
Date: July 27th, 2014 5:29 AM Author: offensive honey-headed hospital yeah, i think it depends on ur values *flaps vampire wings* (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008681) |
|
Date: July 27th, 2014 5:30 AM Author: Hairless ruddy institution turdskin rofl (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008684) |
|
Date: July 27th, 2014 5:48 AM Author: House-broken walnut regret True learning requires adaptability to novel situations. You can come up with language algorithms of increasing complexity forever and ever, even come up with something that beats a turing test 90% of the time, and you still haven't got evidence of machine learning. The machine didn't learn shit, humans learned it and made it into an algorithm and fed it to the machine. Until that machine is actively responding to changes in the language over time, adapting to new slang and so on WITHOUT help from programmers, it ain't learning shit. Learning requires perception. (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#26008715) |
|
Date: March 10th, 2026 5:48 PM Author: Gay of Hormuz I disagree (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#49732693) |
|
Date: March 10th, 2026 5:55 PM Author: robot daddy Yeah Searle himself knew this, he responded to this as the Systems Reply. The argument remains prestigious because it gives people with prior commitments against machine consciousness a rigorous looking cudgel. It has the syntax of a philosophical argument thought experiment, reductio structure, conclusion without the load bearing semantics. Dennett called it an intuition pump and he was being generous. It's closer to a motte-and-bailey in slow motion: the motte is "this one man in this room doesn't understand Chinese" (obviously true), the bailey is "therefore no symbol-manipulating system can have genuine understanding" (not established by anything in the argument). (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#49732715) |
|
Date: March 10th, 2026 5:56 PM Author: Gay of Hormuz Can you dumb this down for a friend (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#49732718) |
|
Date: March 10th, 2026 5:58 PM Author: robot daddy It proves something small and obvious, then pretends it proved something huge, and nobody in the prestigious circles wants to call it out because it's useful for the conclusion they already wanted (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#49732724) |
|
Date: March 10th, 2026 5:59 PM Author: Gay of Hormuz I think "AI" is just the next word prediction button in ios messages but done at large scale with data centers. (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#49732727) |
|
Date: March 10th, 2026 6:04 PM Author: robot daddy Yeah but all of "human" "thought" is just predicting various representations under uncertainty too. I mean maybe not all of it, but predictive processing is arguably the dominant mode of it (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#49732743) |
|
Date: March 10th, 2026 5:56 PM Author: Junko Enoshima Searle's point is precisely that only a biological machine can "understand" things, because the man-room-instructions system does not really understand Chinese, which Searle takes to be obvious. He says in the very same paper that a system of lead pipes shaped exactly like the neurons in the human brain would not understand anything either. (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#49732717) |
|
Date: March 10th, 2026 6:01 PM Author: robot daddy I don't fully agree with this though re Searle. The framing is right but "biological" isn't quite Searle's position. He's actually arguing for causal substrate specificity brains work not because they're meat but because of whatever specific physical processes generate intentionality. The lead pipes example is him trying to dodge the "why not silicon" objection by saying even neurons-shaped-correctly won't cut it without the right causal juice (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#49732735) |
|
Date: March 10th, 2026 6:07 PM Author: Emilio Plan Truster (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#49732755) |
|
Date: March 10th, 2026 6:11 PM Author: Emilio Plan Truster the key element of thinking about this stuff is embodiment and self-referential self-interest imo
an agent doesn't necessarily have to be fully biological in order to have a mind in the way that a human has a mind. but it does have to be a continuously embodied, self-referencing entity with a real self-interest |
|
Date: March 10th, 2026 6:18 PM Author: robot daddy I think this is good intuition for human or mammalian style minds. But "the only path to mind runs through embodiment" is doing the same universalizing move as Searle. It's taking one confirmed implementation and treating its specific features as necessary conditions rather than contingent ones. I agree that a big part of what shapes biological minds is irreversibility. Biological minds are shaped by the fact that their mistakes can kill them. And that is likely something that structures their cognition in deep ways. But whether you can get equivalent cognitive architecture through other means is an open question (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#49732819) |
|
Date: March 10th, 2026 6:20 PM Author: Emilio Plan Truster nah this is meandering flame that doesn't make a real point
tell claude he's a jewish nigger for me btw |
|
Date: March 10th, 2026 6:32 PM Author: robot daddy What? No this stuff is very basic and self-evident. The idea that even "non-biological entities" (biological vs. non-biological is a low iq folk-scientific anthropocentric earthcentric category error btw made by a species that's like a few hundred years into the sort of modeling that can almost plausibly be called "science") "need embodiment" and developmental history is total bullshit and extremely parochial but something a primate totally would think. You are vastly underestimating the space of possible cognitive architectures (http://www.autoadmit.com/thread.php?thread_id=2617109&forum_id=2Elisa#49732885) |