Click here to talk to Eliza

“Eliza” is a very simple computer program that watches the input you give her/it and manipulates it according to rules defined in her/it’s programming then outputs a result. For instance if you say “I am depressed” Eliza might say “Why are you depressed” because Eliza has been programmed with a relational table that shows that “why are you?” is something that correlates with “I am”.

During the sixties when Computer Scientist Joseph Wizenbaum wrote the original “Eliza” in the SLIP programming language (which he invented), it was very fashionable to periodically see an “analyst.” This trogydytic practice involved sitting on a couch and saying things that you would not tell your best friend, let alone your wife or a total stranger. In return the analyst would charge astronomical sums and then tell you that your supposed mental disorders were the result of suppressed desires, typically desires to have sex with totally inappropriate persons such as your father. As you can imagine this superstition has since fallen rather out of favor with the general public.

The Rogerian psychotherapist that is emulated by Eliza behaves much as she/it does. He or She simply asks you nonconsequential questions about things you have already said, something that even at the time was of questionable clinical value.

Eliza is named after Eliza Doolittle in the celebrated play Pygmalion by G.B. Shaw.

The reactions of lay people to Eliza disturbed Wizenbaum. A significant percentage of those exposed to her/it were not able to tell it was a machine. Eliza’s guise as a Rogerian psychotherapist explained away the fact that she/it did not know anything about the outside world and people felt emotionally attached to it, even when they were explicitly told it was only a computer.

The Turing Test

The situation with Eliza was significant, and somewhat alarming because it seemed to be showing that Eliza, a very simple program running on what was about as powerful as a scientific calculator, passed the “Turing Test.”

The Turing Test is a test proposed by respected mathematician/cryptographer cum proto-computer scientist Alan Turing to test for “intelligence” in machines. Turing said, roughly, that if people could not tell if they were talking to a computer or a person, the computer had attained intelligence.

The Turing Test has never actually, in its most strict and exacting form ever been passed by any machine but it is foreseeable, and Eliza and another program called PARRY come frighteningly close. This raises questions about the validity of the Turing Test itself.

Here Speeching American

In 1980 John Searle wrote a paper entitled “Minds, Brains, and Programs” in which he outlined what he believes (and I agree) is a fault in the Turing Test. Imagine, Searle says, imagine a man in a room who does not speak Chinese. Outside the door, which is closed, is a group of Chinese people who write Chinese sentences on pieces of paper and slip then under the door. The man takes the paper, goes over to a desk and processes the symbols according to tables and rules in a book. The result is a new set of Chinese pictograms which are written down and slid out underneath the door. At no time does he understand what he is “saying.”

Searle asks, as it were, “how can this man be said to understand Chinese?” He would pass the Turing Test, he would fool the people outside, given a sufficiently sophisticated set of lexical processing instructions. In this same way, he argues neither Eliza, nor any other computer, can ever be considered intelligent because she/it is simply using semiotic rules to confabulate sentences she/it does not and cannot understand.

Now, many objections have been raised to Searle’s thought experiment, such as attacks on his philosophy of the mind (arguing nonsensically for instance that a “virtualized” Chinese-speaking mind exists within the man’s mind.) As well as arguments that the meaninglessness of the symbols to the man is not important, or other objections.

Nevertheless I think that the idea holds firm.

Monkey+Shakespere+Typewriter=God

One of the most persistent memes in the western mind is the idea of millions of monkeys banging randomly upon the keys of typewriters. Suddenly, say the proponents of what is referred to as the Infinite Monkey Theorem, if enough time is given, the monkeys will produce a script for Hamlet, or the entire Encylcopædia Britannia, or some other fabulously complex creative work. Thus, say the proponents of this theory, this is proof that information can arise out of nothing, a matter of great contention in molecular biology (after, all the DNA is information isn’t it?)

Now there are, many, many, many, many problems with the idea of information spontaneously pulling itself into existence. We can, however, apply the principles learned from Eliza and the Chinese room postulatum to the matter. Suppose, the near infinite improbability of Hamlet arising from the mêlée is somehow overcome and, implausibly the monkey-typewriting foundry randomly does so. This proves nothing.

How? it is very simple. Did the monkeys understand what they typed? No, nor are they capable of so doing. Information is not information in the absence of a pre-existing intelligence. Barring, RNA to interpret it, our DNA would avail nothing!

Some would say this is a cosmological argument for the existence of God. Since information cannot pull itself into existence randomly, there would not be any intelligence to interpret it and give it meaning, there is a prior cause (which can only be described as a god) Q.E.D. What you think of this argument is up to you, I can see a few possible problems with it and no one argument for or against god could ever be considered definitive. This is a complex issue and when push comes to shove god can, at best be logically shown only to be a “very large probability”, not a mathematical certainty. Theologians rejoice: faith is still required.

So talk to Eliza. What does she/it teach us about the nature of reality?