Let’s say I have a computer program that speaks Chinese as well as a human native speaker of the language; I’ve had people who speak fluent Chinese talk to the program using instant messages for an extended period of time, and none of them have been able to figure out that my program isn’t another person. Now, suppose I print out an English version of this program and store it in some filing cabinets in a room. I go into the room, lock the door, and begin exchanging messages with a Chinese speaker by passing notes under the door. Each time a note is passed in, I follow the written steps of my program to figure out what reply to make, write that down, and then slide it back under the door. Like before, anyone reading the notes I pass back under the door is completely unable to tell that I’m not a native Chinese speaker.¹

John Searle formulated this thought experiment so that he could ask whether the person in the room could be said to understand Chinese in the same sense as a native speaker. If we admit that they are not the same, as it seems we must, we must also admit that a computer running the program likewise does not understand Chinese, and thus that it is impossible for a computer to think and understand as humans do.

The problem with this conclusion is that it requires that neurons and brains possess qualities which transistors and computer chips do not. Searle suggested that neurons have “causal properties” which cannot be detected by outside observers but nevertheless give rise to the mind and consciousness. Another possibility is that, as Sir Roger Penrose has proposed, microtubules in neurons exploit quantum effects to perform hypercomputation.

I would like to propose that brains are actually made out of magical fairies, which is equally well-supported by the available evidence and easier to explain to young children.

Or perhaps it is time to go back and double-check the logic leading to Searle’s conclusion.

The crux of Searle’s argument is that brains do this thing called “understanding” that computer programs (algorithms, to be more precise) cannot. If we are reductionists (and we should be!), then understanding is something that can be explained in entirely physical terms. Either there is an as-yet-undetected Fairy Force at work in human brains, or the Chinese room is capable of understanding.

There is no choice but to bite one of these bullets; I find the latter more palatable.

 

1: Except for how long it takes for me to make my replies. We can ignore this for the sake of the thought experiment, but I should still note that it would take an unimaginably huge amount of time to walk through every step of a complex computer program that was designed to run on a machine that performs 2 or 3 billion operations each second.