JOHN R. SEARLE'S CHINESE ROOM

John Searle
Image Courtesy:faculty.washington.edu

The Chinese room Experiment is a pathbreaking experiment in the field of Artificial Intelligence.This thought experiment by John Searle which first appeared in his paper "Minds, Brains, and Programs",published in Behavioral and Brain Sciences in 1980.[1]It answers a major question that if a machine is able to carry on a normal intelligent conversation in a language with a human being,does it necessarily understand that language?

The Idea Behind The Experiment

When can we say a computer is as intelligent as a human being? The question we raise here is not about the quantitative measure of intelligence, but rather a qualitative one. Can we build a computer which really understands things in a fashion we do ?

For an insight into what we are trying to get at, let us consider the following instructions:

Step 1: X = ith number in the list.

Step 2: if (X > Y) Y = X

Step 3: Do for i = 1 to n

Step 4: Output Y

Now this program can be executed easily on any capable machine. But does the machine which is executing this code (the computer in this case), knows what is it doing ? Does it know that we are actually trying to compute the maximum among all numbers in the list. Well, we know computers just execute instructions, they may not know what they are doing.

This raises a valid concern, machines may produce a desired result, without actually knowing or understanding of what they are doing as our mind has. So can we call these machines intelligent ?Or seen from a different perspective can we say that machines have a mind of their own ?

Do get a deeper insight consider the problem of building a system which can predict weather. Now suppose we come up with such a system with predicts weather with great accuracy. So we have a black box which takes as inputs say the temperature, the wind speed, the humidity etc as inputs and outputs the weather prediction. Now does this black which makes predictions, actually capture the way in which weather really changes or is it just a model?

Stemming from these 2 viewpoints, we have 2 schools of thought. One which says the simulation is the actual system (Strong AI), while the other says the simulation is just a model of the actual system (Weak AI).[2]

The Chinese Room Experiment is based on the disapproval of the former school of thought,ie.Strong AI which is "The Artificial Intelligence that matches or exceeds human intelligence � the intelligence of a machine that can successfully perform any intellectual task that a human being can. [2]

The Experiment

Chinese Room

Image courtesy:alexandria.nu

Now, suppose I have a computer which takes into inputs some Chinese characters and gives as an output some Chinese characters. Over here we are essentially concerned with a model of computation based on symbol manipulation, i.e given as input a set of symbols we do some operations on thisset to produce as result. Now, a Chinese asked to converse with this computer (say on a chat client) without revealing to him whether he is Given this, can we say that computer has a mind like us ?

Now suppose we make a hardcopy of the program the computer was executing, and replace the computer by a human who does not understand Chinese at all. Searle says that there is no major difference between the role played by the computer in the first case and the role played by him in the second case. Both are simply following a program, step-by-step, which represents intelligent behavior.It follows that this human should be able to give the same output as the computer by executing the code manually. Thus, apparently the man seems to be understanding Chinese. But we know he does not.This disproves the notion of Strong AI


Video Courtesy:Youtube

Conclusion

The Searle's Chinese Room Experiment concludes that a program cannot give a computer a "mind" or "understanding", regardless of how intelligently it may make it behave.It says that programs are neither constitutive of nor sufficient for minds.The computer can only follow programs,it cannot think and modify programs on its own.[1]

References

[1] Minds,Brains and Programs,published in Behavioural and Brain Sciences in 1980:The paper which presents the Chinese Room Experiment

[2]An Article on the book Singularity is Near by Ray Kurzweil:The book that brought the notion of Strong AI