Colorless Green Ideas Look Like THIS!

Colorless Green Ideas Look Like THIS!


It was impossible to imagine this picture - until DALL-E did it.

The linguist Noam Chomsky (7 December 1928 - ) gives us a famous sentence that was supposed to be impossible to imagine - that is, until DALL-E came along.

In his 1955 thesis, The Logical Structure of Linguistic Theory, the renowned linguist was trying to demonstrate the difference between syntax and semantics.

Syntax analyzes language at the structural level; semantics is concerned with the level of meaning.

Essentially Chomsky wanted to show that you can have a sentence that is perfectly grammatical (i.e., following the rules of syntax) while still being non-sensical (i.e., breaking the rules of semantics).

To do so, Chomsky coined the following now-famous sentences as an example:

(1) Colorless green ideas sleep furiously.

(2) Furiously sleep ideas green colorless.

Try this: read each sentence out loud.

Did you notice anything different in the way that you read those sentences? Chomsky predicts that a native English speaker would read (1) with the ordinary intonation pattern, whereas he would be forced to read (2) "with a falling intonation on each word, as in the case of any ungrammatical string."

Why is that? Why do we read these sentences differently?

We read (1) with ordinary intonation because we recognize it as being grammatical. In other words, (1) follows the syntax of English sentence structure, which we see below.

Chomsky's sentence is perfectly grammatical, yet we still can't imagine it; it doesn't mean anything.

By contrast, we recognize that (2) is ungrammatical because the words don't conform to English syntax - i.e., like placing a square peg in a round hole, we couldn't fit (2) into the structure illustrated above. And because it doesn't fit into this normal syntax structure, we can't read it aloud with the same intonation patterns that we use for grammatical sentences.

Chomsky's point here is that we can have grammatical sentences that don't mean anything.

You know that (1) is grammatical; however, if I gave you a paint brush, would you be able to draw it?

DALL-E is an A.I. program that generates images from writing prompts. Developed by OpenAI, DALL-E has a sophisticated model of the relationship between images and the language used to describe them. It can thus produce visual representations of natural language.

For example, when I gave it the sentence "My hovercraft is full of eels," a nod to a hilarious skit about bad English translations from the British comedy show Monty Python, here's what it produced:

"My hovercraft is full of eels." –DALL-E (2022)

So I wanted to see what DALL-E would come up with when I gave it Chomsky's famously unimaginable sentence.

An unimaginable sentence imagined by A.I.

Not bad, I must say.

More posts in linguistics