Wednesday, July 15, 2009

Alan Turing's brilliant essay

In 1950, Alan Turing wrote "Computing Machinery and Intelligence." This one short paper, exploring what came to be called the Turing test, continues to influence research and thinking across multiple fields.

Tyler Cowen and I have co-authored a new paper asking two questions. What does the Turing test really mean? And how many human beings (including Turing) could pass? Our premise is that some aspects of Turing's paper have not received sufficient attention:

Turing’s paper is rich and multi-faceted and we are not seeking to overturn all of the extant interpretations. We do wish to suggest that a potent and indeed subversive perspective in the paper has been underemphasized. Some of the message of Turing’s paper is encouraging us to take a broader perspective on intelligence and some of his points are ethical in nature. Turing’s paper is about the possibility of unusual forms of intelligence, our inability to recognize those intelligences, and the limitations of indistinguishability as a standard for defining intelligence. “Inability to imitate does not rule out intelligence” is an alternative way of reading many parts of his argument. Turing was issuing the warning that we should not dismiss or persecute entities which we cannot easily categorize or understand.
The facts of Turing's life enter into our argument, as does autism in many respects. Here is what we conclude:

It is possible that Turing conceived of his imitation test precisely because he had so much difficulty “passing” and communicating himself. In social settings these facts were seen as disabilities but in the longer term they helped Turing produce this brilliant essay.
Tyler Cowen is a professor of economics at George Mason University. His page is here. He blogs at Marginal Revolution; his post about our paper is here.


abfh said...

A very interesting analysis, indeed. I have become much more aware, through my involvement with online discussion in the autistic community, of the limits of words and the many different ways in which they are used. Even so, I still have to remind myself sometimes to give people the benefit of the doubt when they use words in unusual ways.

I'll be buying Tyler Cowen's new book the next time I go to the bookstore.

Catana said...

Wondrous serendipity! A very thought-provoking post, plus Tyler Cowan, who I discovered just a few days ago when someone posted the link to his Chronicle article: Autism as Academic Paradigm. I've thought a great deal and written a little about the variety of minds and the difficulty of recognizing anything that's too different as equally valuable or maybe even superior to the norm.

daedalus2u said...

This is very interesting, and it parallels much of my own thinking. I think that imitation (or a term I like better emulation) is absolutely essential for all communication.

I agree with Turing that the concept of “can machines think” is somewhat nebulous to discuss, but that any such question necessarily involves ideas being communicated to and from the entity being tested for “thinking”. Communication is an under appreciated aspect to the imitation game.

In a very real sense, the only things that can be communicated are mental states. The concept being communicated must first exist in the mind of the first individual, be translated into a data stream of a communication medium, the data stream must be transferred and then the data up-converted back into the mental concept.

If the two individuals do not share the same communication protocol for the conversion or mapping of the communication data stream into the mental concept, then the mental concept cannot be communicated. This communication is like a chain, it can be interrupted at any link. If the communication is interrupted, there is no way to tell where in the chain it is interrupted.

I discuss some of this in a recent blog.

In rereading Turing’s article, I agree that there is a lot of stuff between the lines that he couldn’t say in 1950. I think the idea that a gay man should be respected as an individual with human rights was a mental concept that most people in his time couldn’t begin to understand. I noticed a line in discussing educating a “machine” which perhaps was autobiographical, ”Possibly it might not have eyes. But however well these deficiencies might be overcome by clever engineering, one could not send the creature to school without the other children making excessive fun of it.”

I think that when people communicate with each other, they do an implicit “Turing Test”, to determine if the other individual is “human enough”. If someone fails that Turing Test, then xenophobia is triggered, which results in the bullying that Turing mentioned regarding school children and which (I think) ultimately took his life.

laurentius rex said...

Turing Schmuring, poor old sod, history decreed (determinism that is) that he could never hear of the motor thery of language never mind the cognitive linguists and never mind the deconstruction of sacred mathematics from those who not so much see them as the equivalent of a "language game" but an inevitable consequence of the human condition of our particular embodiment and cerebral heritage. Think yourself outside of the cage and the question becomes, not so much (as is often raised by Turings ideological descendants) whether that kind who can think such is human but whether any other system is possible.

I dare to think the impossible, but then I am not sure whether I really have that choice or not, or that some deep determinism is describing and inscribing my words as I think I write.

joel said...

This is my field - there are several problems with the idea of the Turing test, the biggest being that a subset of humans fail it when it is researched. (the other problem is that things pass it which do not meet the intent)

Turing100in2012 said...

@ joel That's quite proprietorial "my field"! :-) Turing's tests (both three-participant, and one-to-one scenarios) encompass a wide area.

What do you mean by "a subset" of humans fail "when it is researched"? Are you pointing to instances in the Loebner Prize?

Interrogators mistaking a human for a machine (the confederate effect), in a contest situation actually says a lot about what humans consider machine-like talk to be, and it varies from judge to judge during instantiated Turing tests. (See Shah & Warwick's upcoming paper in Kybernetes: Testing Turing's Five Minutes Parallel-paired Imitation Game)

Muthu Pearl said...

beautiful essay of Alan Turing .