Language and Thought

For as famous as it is, I don’t think most people understand the Turing Test. People seem to know that:

• A computer has a textual conversation with a human.
• The computer attempts to trick the human into thinking it isn’t a machine.
• Turing said, if that computer succeeds, we should call it intelligent.

Unfortunately that’s as far as most people go. They don’t seem to ask themselves why Turing thought that the Imitation Game defines intelligence.

Most seem to imagine chatbots like Weizenbaum’s ELIZA:

ELIZA: Your boyfriend made you come here?
Person: He says I’m depressed much of the time.
ELIZA: I am sorry to hear that you are depressed.
Person: It’s true. I am unhappy.
ELIZA: Do you think coming here will help you not to be unhappy?
Person: I need some help, that much seems certain.

“Neat, it sounds just like a human! It’s obviously not very intelligent though… That Turing guy must have been wrong.”

Let me be clear here: the Turing Test has nothing to do with being able to hold a conversation.

This week I’m writing about a different test — it’s a kind of “Mini Imitation Game.” But before I do that, some might need a taste of the original Imitation Game. This post is a short introduction to the Turing Test, getting at the question “why did Turing consider the Imitation Game a suitable definition of intelligence?”

Why The Imitation Game

When Turing wrote Computing Machinery and Intelligence, he started with a now almost cliché question: “can machines think?” This would lead him to two ideas — one about intelligence, and another about language.

Intelligence

On the intelligence side of things, Turing was concerned by the fact that “to think” is a pretty undefined term. In fact, we almost treat it as a matter of opinion. Do I think? Do you think? Does my dog think? We tend to tie up the word with some idea of a soul, or another by-definition-undefinable quality.

Turing’s solution to this problem was to avoid it entirely.

Turing’s Definition of Intelligence

Anything that can behave like a human must be at least as intelligent as one.

If it behaves like it’s intelligent, who are you to say it isn’t? Who cares what’s going on inside? Turing is saying that looking into the black-box of intelligence isn’t necessary to know something can think — the only thing that’s important is how something acts.

But Tyler, aren’t there intelligent things that can’t behave like humans?

I don’t know, and neither does Turing! All he is saying is there aren’t things that can behave like humans which aren’t intelligent.

\[ \text{Behave Thinky} \subseteq \text{Thinky} \]

Language

Well… then how do humans behave?

To answer this, Turing had a pretty powerful hot take:

Turing’s Observation on Intelligence

There aren’t tasks/behaviors which humans do, about which they cannot speak.

Describe one thing we do that is indescribable… It’s OK, I’ll wait. While I’m waiting, let’s make clear some things a human could speak about:

Not a single thing on this list we would call a language skill — instead we have math, problem solving, hypothesizing, and empathizing. Turing wasn’t picturing asking a computer questions like “how are you today?” Instead, he was thinking about questions like “what would go wrong if I made my desk out of bacon?”

The Turing test isn’t a test about the mastery of language. It’s a test about the mastery of thinking, expressed through the universal medium of language.

Language and Thought

“The universal medium of language” is an incredibly powerful idea, central to the Turing test, and much modern thought. The idea is pretty simple: you can’t name something you can’t talk about, because if you did, then you’re talking about it.

There are many ideas out there about the relationship between language and thought. Linguistic Determinism and Linguistic Relativity are strong and weak versions of the idea that language is thought. Language of Thought theories argue that expressed language is not essential for unexpressed thought. Chompsky’s independent theory argues that language is one cognitive system, working in concert with many others. But in all of these cases, we run into the same conclusion: language is our only way of expressing our own cognition. To quote Chomsky:

Language is a mirror of mind in a deep and significant sense. It is a product of human intelligence… By studying the properties of natural languages, their structure, organization, and use, we may hope to learn something about human nature; something significant, if it is true that human cognitive capacity is the truly distinctive and most remarkable characteristic of the species.

What Did Turing Picture?

Turing was clear that his game was a high bar for intelligence. He considered passing to be sufficient, albeit not necessary, in saying something is intelligent. And he went out of his way to defend the power of his definition, even addressing how the test would have to change for humans with psychic abilities (yes, that’s real).

To wrap things up, here’s a now famous excerpt from Turing’s original paper. I think it gives us a flavor for what he was picturing when he wrote about the Imitation Game.

Interrogator: In the first line of your sonnet which reads “Shall I compare thee to a summer’s day”, would not “a spring day” do as well or better?
Witness: It wouldn’t scan.
Interrogator: How about “a winter’s day.” That would scan all right.
Witness: Yes, but nobody wants to be compared to a winter’s day.
Interrogator: Would you say Mr. Pickwick reminded you of Christmas?
Witness: In a way.
Interrogator: Yet Christmas is a winter’s day, and I do not think Mr. Pickwick would mind the comparison.
Witness: I don’t think you’re serious. By a winter’s day one means a typical winter’s day, rather than a special one like Christmas.

Could it be that there are intelligences out there which couldn’t communicate like this? Quite possibly. But if you were to claim that the Witness in this dialog wasn’t intelligent, that says a lot more about you, than it does about the Witness.

Reading

• Computing Machinery and Intelligence, Turing’s paper which introduces the Imitation Game, is an incredibly compelling read. I feel almost silly for having summarized it.
• For more context on Language and Though, I’d recommend A Review of the Language-Thought Debate: Multivariant Perspectives, by Parviz Birjandi and Somayyeh Sabah.