All replies to the Chinese Room Argument (CRA) that I've seen assume the computer science concept (explanation, "definition" as Searle says) of the electronic digital computer. But what of other conceptions? If there are alternative understandings that are true to the hardware and also could allow the acquisition of an internal semantics, would this rebut the CRA? Even if it is not clear that an alternative conception allows inner semantic content, does the fact that (a) such an alternative exists, and (b) that the CRA has not addressed it, rebut the CRA? For wouldn't the CRA's conclusion then merely be that the computer science concept (i.e., Turing's concept) of the machine could never think whereas other conceptions that are true to the hardware, might?

  • CRA is specifically directed at the "computer science conception" (so-called "strong AI") not being able to explain "understanding", as Searle sees it, not at the generic idea of "thinking hardware". So all alternative conceptions of it are moot to his argument. Searle himself allows that we may well be able to build thinking machines one day, he just thinks that it takes material with special causal properties (like organic matter) for these machines to attain "understanding", see Biological naturalism.
    – Conifold
    May 10, 2019 at 0:49
  • @Conifold And Searle says humans are thinking machines. Suppose there's an alternative concept of the computer that is accurate, and there are reasons to suppose this other conception allows an internal semantics, would this be a rebuttal of the CRA? If not, and if there is such other conception then supporters of the CRA would still have their argument, and AI would still have its machine. AI would need the better concept of the machine but that's a much better outcome than the machine being forever a prisoner in a universe of mere syntax. Does this idea, sort of a win-win solution, stack up?
    – Roddus
    May 12, 2019 at 3:53
  • No, it would not be. Arguments have limited scope, they can not rule out vaguely anticipated somethings that may one day emerge, such as conception of a machine which is we know not what. It is possible that CRA can be modified to argue against such conceptions as well, but we'd have to see what those conceptions are first. Arguably, even artificial neural networks are sufficiently different from the von Neumann computers Searle was caricaturing to raise doubts about intuitive persuasiveness of CRA against them, upon which he relies. Syntax is Semantics is the motto of Rapaport's programme.
    – Conifold
    May 12, 2019 at 10:40

1 Answer 1


John Searle takes the Chinese Room Argument beyond the Turing machine model in the "brain simulator reply" in section III in Minds, Brains and Programs:

The problem with the brain simulator is that it is simulating the wrong things about the brain. As long as it simulates only the formal structure of the sequence of neuron firings at the synapses, it won't have simulated what matters about the brain, namely its causal properties, its ability to produce intentional states.

Later he remarks that "formal properties are not by themselves constitutive of intentionality". These formal properties are not restricted to Turing machines.

Searle, J. R. (1980). Mind, brains and programs. A debate on artificial intelligence. The Behavioral and Brain Science, 3, 128-135.

  • I don't really understand Searle's response. He talks about water in pipes simulating neural pulse propagation. Why not dispense with water molecules and replicate the brain organic molecule for molecule? Would he still say there is no intentionality? I take the Church-Turing idea of simulation where a system is quite accurately described and the description run through a computer (as the program). Then the output of the program describes what would have been the output of the system, had it received as input what is described by the input to the program. I'm just confused.
    – Roddus
    May 12, 2019 at 5:01
  • @Roddus What he is suggesting with the water pipes is that if one simulates understanding by some other means than a Turing machine, he could still put in a human to simulate the formal structure of the simulator and that human would still not understand. The water pipes is just an example of another kind of simulation of neural activity. What has to be done is not simulate the formal structure but somehow reproduce the causal properties of understanding. What makes the problem difficult is that it is unclear what those causal properties are. May 12, 2019 at 9:25

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .