2

I'm a programmer interested in the development of artificial intelligence. I just read about the Chinese Room argument by John Searle, and would like to know if my reasoning successfully refutes his argument.

The Chinese Room argument more or less points out that a computer program might simulate consciousness / sentience, but never actually experience it in the same way a man, ignorant of the Chinese language might simulate knowledge of the Chinese language via a set of instructions, without ever actually experiencing the ability to understand Chinese himself.

I can see how Mr. Searle would arrive at his conclusion, but I believe he made an error.

An AI, perhaps Apple's Siri, upgraded 20 years into the future with additional capabilities would be programmed via algorithms, instructions, to interpret information and respond to stimuli, demonstrating without actually experiencing the vast complexity of thought patterns that form sentience, as Searle argued through the Chinese Room.

However, consider a different form of this AI simulation, one that perfectly mirrors the human brain:

What if we were to invent a machine capable of scanning and recording every cell, every electron, every velocity of every atom in a human brain, and then feed that data into a computer powerful enough to simulate that entire brain's existence computationally?

This simulated brain, as a modifiable data structure, would be a powerful AI capable of developing beyond the physical limitations of a physical human brain. The brain, regardless of it's medium for existence (a computer simulation), would be completely blind to the fact that it is in a computer simulation at all. Yet, it would perform in a perfectly identical manner to the brain of the person scanned, assuming all of the proper physics and stimuli are present in the simulation.

Therefore, who is to say that this AI, this simulated brain, is any less sentient than you or I?

  • And by the way, I originally asked this after reading somewhere on this site that the Chinese Room hadn't been refuted validly yet, but I now see there's a question here that covers various ways it's been argued against. I think my question is still valid though. – Viziionary Jul 14 '15 at 19:37
  • 5
    The Chinese room argument is an argument to the effect that simulating our brain wouldn't be enough to produce conscious experience. You're suggesting that we could simulate our brains. How does that refute the argument? – Quentin Ruyant Jul 15 '15 at 9:23
  • 2
    Chinese room is an argument against "strong AI". According to Searle "strong AI" promises certification of "mental states" beyond what we ask even of other people, from inside the room so to speak. Imitating human brains therefore does nothing, not being able to give a detailed account of how "mental states" emerge misses the point and already concedes "strong AI". Searle's weak point is the reification of "mental states" and "meanings", not the imitation issues. – Conifold Jul 16 '15 at 19:10
  • 1
    The Chinese room argument is irrefutable for exactly the same reason solipsism is irrefutable -- there's no way to prove that anyone else's mind exists either. The fact that other people appear and claim to be sentient is not evidence of the fact that they are; likewise, duplicating a human brain with technology might produce something that seems to be sentient, claims to be sentient, etc., but it would never be able to prove to John Searle that it is sentient, even if it were a duplicate of his mother. It's an absurdist form of skepticism. – selfConceivedAsEvil Aug 17 '15 at 12:22
4

The short answer

No, sorry this does not refute the Chinese Room argument.

A more detailed answer

In this question I briefly talk about the hard problem of consciousness, and how it is not a question of what, but a question of why.

Having a perfect diagram that in real time mimics the movements of neurons of a brain, say, a brain in a vat would be interesting for sure. But the simulation and the one in the vat, though identical in structure (though one is nonphysical, at least in the sense that it is not a brain, but circuitry) have a crucial difference.

That difference is consciousness, qualia, experience. Your hypothetical AI relies on some answer to the hard problem of consciousness. Otherwise it is just like a movie, playing the same movements as the brain in a vat, but having no actual thoughts. Much like the philosophical zombie thought experiment.

  • 1
    What? The mind created by the simulation thinks no differently from the brain in your head. Who's to say that's any different from the possibility that our own universe is a computer simulation? Might we then, too be philosophical zombies? Why or why not? – Viziionary Jul 14 '15 at 20:13
  • @MediaWebDev So a simulation of the movements of neurons is not equal to the thought that the neurons create. In order for them to be the same the hard problem of consciousness would need to be solved. That is not a trivial task. – hellyale Jul 14 '15 at 20:39
  • 1
    You claim that the outcome would be similar to a philosophical zombie, but that is defined as something without sentience. Do you even understand anything about neurology?? The brain isnt magical. Sentience isnt magical. It's nothing more than a very highly complex set of patterns and reactions between small communicating mechanisms - whether those mechanisms take the form of neurons or objects in a computer program does not matter - the sentience is real nonetheless. – Viziionary Jul 14 '15 at 22:22
  • 1
    @MediaWebDev Chalmers wouldn't agree, but that is ok. Copying the brain is not an answer to the hard problem. Even if it was possible, it does not answer the problem. It is not a question of how, it is a question of why. Regardless your answer is not a refutation of the Chinese room. In no way does it explain where the experience comes from. Your claim just assumes that the copy will have consciousness, but does nothing to explain why. – hellyale Jul 14 '15 at 22:41
  • 1
    @Viziionary If you simulate a hurricane, nothing actually gets wet. Consciousness may be like things getting wet. – David Schwartz Jun 1 '16 at 17:42
3

At the end of "Can Machines Think?" Searle writes:

"For any artifact that we might build which had mental states equivalent to human mental states, the implementation of a computer program would not by itself be sufficient. Rather the artifact would have to have powers equivalent to the powers of the human brain."

The Chinese Room Argument is supposed to show that minds are not just really complex computer programs. If a computer can have the powers of a human brain in virtue of its implementing a really complex computer program, then Searle would agree that this computer is a mind.

  • Perhaps Searle, in all of his philosophy might agree that it is a mind, but I see no way for his Chinese Room thought experiment to agree. Perhaps then, the Chinese Room is refuted by my reasoning, but not Searle in all of his composite thinking on the subject? – Viziionary Jul 14 '15 at 22:26
  • Just read Can Machines Think?, it's available online and he explicitly deals there with the Chinese Room Experiment. The Chinese Room thought experiment is more modest than you think; it's not supposed to show that you can't possibly create a machine that understands language, it is only meant to show that pure syntax is not enough for understanding. – rds Jul 15 '15 at 1:21
  • 1
    For some reason Searle is committed to "mind" coming from biochemical implementation of the human brain. So I am not sure that computer having the powers of a human brain by implementing a really complex computer program is enough for him. He also insists on some sort of manifested "mental states" in addition to just simulating the "powers". – Conifold Jul 16 '15 at 18:58
  • Searle accepts multiple realization; even if he says that the human mind comes from biochemical interactions, he can't say that only biochemical reactions can cause a brain. Also, Searle claims that mental states are not pure syntax, but this doesn't mean that you can't create a mind; it means that if you create one by implementing really complex code, her mental states can't be reduced to pieces of code. Finally, I don't really know Searle ouvre extensively, so if you can say where does he talk about "manifested" mental states I'd be grateful. – rds Jul 16 '15 at 20:57
3

First, as Luís Henrique pointed out, the physical feasibility of this computer is debatable at best, but for the sake of argument, lets assume that it is.

Level 1 answer

No, your reasoning does not refute the Chinese Room argument. Underlying the Chinese Argument is Searle's more fundamental principle that syntax is not semantics. Your super-powerful computer, for all of its computational and memory capabilities, is still manipulating abstract symbols, and it does so without knowing the meaning of those symbols. The millions of individual data points it is manipulating can be the electrons and nuclei of a human brain, the planets and stars in a galaxy, or the individual citizens of China,...all your supercomputer has is a set of symbols and rules for manipulating those symbols, and it doesn't know or care what they mean. Hence argues John Searle, it can never be truly conscious or sentient.

Here's another way of looking at it: A computer performing a completely perfect simulation of a flying bird (in terms of all the physical data concerning the bird's velocity and position, mass, air resistance, etc...) is still just preforming a simulation, it doesn't constitute an artificial recreation of bird flight.

Level 2 answer

You are pointing out a weakness in John Searle's "syntax is not semantics" principle. John Searle himself admits in his lectures (sorry I forgot which lecture, you will have to go through at least the first 6) that a biological simulation of the brain might be conscious in a way that a digital computer simulation wouldn't be.

But that begs the question: Why this barrier between the biological and the digital? Why can't a digital electronic recreation of individual neurons then interconnected to form an artificial brain be conscious? To use the bird analogy from above: A perfect computer simulation of a bird flying might be a mere simulation, not real flight, but a flying robotic bird (or an airplane) is undeniably flying. In the same sense: Why can't a computer equipped with AI algorithms be equivalent to a human brain in the same way that a plane flying is equivalent to a bird flying? Or that a prosthetic leg is equivalent to a biological leg?

This position, that it is enough for a computer to perform the same functions as the human mind/brain is known as functionalism. Searle was trying to refute functionalism, and the question one could throw back at him: You say that syntax is not semantics, but if the semantics are not in the functional description of the mind, where else can they be? The semantics can only be in the functional aspect of the of mind. Whoever wrote the rules and instruction for the Chinese room, did so using the semantics of the Chinese language, and therefor those semantics are reflected in the syntax, and syntax and semantics are not as separable as Searle believes they are. See this answer, this answer, and this answer to my above mentioned question on syntax vs semantics for more detailed (and eloquent explanations).

  • "Your super-powerful computer, for all of its computational and memory capabilities, is still manipulating abstract symbols, and it does so without knowing the meaning of those symbols." Are you familiar with Simulation Theory? By that logic, none of us can claim sentience. Actually, it's that kind of absurdist skeptisism that another comment explains in the basis for the whole argument. When I asked the question, I didnt realise this was a theory of "nothing is real". Seems pretty silly. – Viziionary Sep 21 '16 at 2:02
  • @Viziionary did you read the rest of my reply and the links within it? – Alexander S King Sep 21 '16 at 2:13
1

What if we were to invent a machine capable of scanning and recording every cell, every electron, every velocity of every atom in a human brain, and then feed that data into a computer powerful enough to simulate that entire brain's existence computationally?

One problem is that such machine is impossible; it violates Heisenberg's principle.

1

Stating that Searle's Chinese Room Experiment (CRE) hasn't been refuted is, in my humble opinion, somewhat of an understatement. The conclusion from the CRE is tempting but we can clearly state that the CRE is faulty. You just have to read the comments of Douglas Hofstadter (Hofstadter and Dennett - the Mind's I (2001)), Daniel C. Dennett (Intuition pumps and Other Tools for Thinking (2013) or Jack Copeland (Artificial Intelligence (1993). They all state that the CRE is faulty by design and therefor cannot state that a machine (computer) cannot think. You asked if an AI could have consciousness but that wasn't the question of the CRE. The (original) CRE was a method in which Searle stated that no machine could think (in this, could pass the Turing Test). Jack Copeland and Daniel Dennett both wrote that it's philosophically possible to speak about machines that can think. Dennett on the basis of his intentional stance argument. Copeland states that when a machine (or whatever) can think if this machine can do exact the same processes as a human does when we speak of thinking. On the basis that when we use 'thinking' for behavior of a human being, we can also use the same word for the same behavior of a machine (read chapter 3 of his book mentioned before).

Untill now there hasn't been a machine that met the requirements of the Turing Test so one could state that there hasn't been a machine yet that can think.

  • Good references, but would you have page numbers so the reader can more quickly find the specific parts in those books about the CRE. Regardless, welcome to this SE! – Frank Hubeny Nov 22 '18 at 16:52
  • "Searle stated that no machine could think (in this pass the Turing Test)." -- Oh my goodness, no. The Chinese room DOES pass the Turing test (at least with respect to speaking Chinese). It's indistinguishable from a native speaker of Chinese. Yet it is not self-aware at all. The Chinese room argument (usually CRA in the literature) shows the difference between passing the Turing test and being conscious. My neighbor passes the Turing test. I have no way of proving he's conscious. – user4894 Nov 22 '18 at 19:15
  • @ Frank Hubeny: For ‘the Minds I’ see Hofstadters reflections following Searle’s article ‘Mind, brains and programs’. In this reflection Hofstadter point out what is faulty about the CRE. For Dennett see chapter 60 ‘The Chinese Room’. For Copeland’s see chapter 3 ‘Can a machine think?’ and chapter 6 ‘ ‘The curious case of the Chinese room’. You might also want to read Georges Rey article ‘Searle’s Misunderstanding of Functionalism and Strong AI’. You can find it here: sites.google.com/site/georgesrey. – Pascal Nov 23 '18 at 15:58
  • @ user4894: The Chinese Room is not the same as the Turing Test (TT). Searle, in his original article ‘Minds, Brains and Programs’, wanted to proof that the TT does not proof that a machine can think and therefor that strong AI does not exist. It’s about intelligence and intentionality. Later on, the discussion broadened to consciousness. Intentionality and consciousness are two different things. Something can have intelligence and intentionality but at the same time lack consciousness. – Pascal Nov 23 '18 at 16:05
0

The argument is from an analogy between Chinese and consciousness. So you haven't shown that the argument is based on a mistake, if you are indeed just arguing that:

  • a very complex computer could simulate everything that makes up a human mind

as you seem to.

The only way I can imagine you have begun (assuming no error in validity any argument isn't going to be utterly disposed of) to refute Prof. Searle is that you're asserting either:

  • a very complex computer is beyond our current human reasoning,

or

  • if we simulate every aspect of Chinese then we do understand it.

Maybe you are blurring together these two claims, which would not be very helpful. I'm sympathetic to both claims, but you wouldn't convince Prof. Searle.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.