Emily Bender Explains What 'Stochastic Parrots' Really Means, Five Years Later

In March 2021, four linguists and artificial intelligence specialists published the article "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?", which introduced the metaphor of the "stochastic parrot" into circulation — a system that statistically predicts words without understanding meaning; five years later, the article's lead author, Emily Bender, a professor of computational linguistics at the University of Washington, published a post breaking down the main misconceptions about the term and discussed this in an interview with IEEE Spectrum.

Why the article made such a stir

Attention to the article was fueled by a scandal: shortly before its publication, Google fired two of its co-authors — researchers Timnit Gebru and Margaret Mitchell. The work argued that large language models generate text by statistically predicting probable sequences of words, rather than understanding what was said — hence the metaphor of a "parrot" repeating patterns without understanding.

• The article was published in March 2021 and has four authors.
• Google fired co-authors Timnit Gebru and Margaret Mitchell shortly before the publication.
• Over five years, the metaphor went far beyond academia and inspired projects like the shoulder-mounted robot "Stochastic Parrot," created at MIT Media Lab.
• On the fifth anniversary of the article, Bender wrote the post "Stochastic Parrots: Frequently Unasked Questions" to dispel misconceptions about the term.

How the metaphor went beyond academia

Originally, the "stochastic parrot" was a term used by a narrow circle of computational linguistics and AI ethics specialists. But over five years, the metaphor penetrated mass conversation about artificial intelligence so thoroughly that it became the starting point for projects far beyond linguistics — including the aforementioned shoulder-mounted robot "Stochastic Parrot" from MIT Media Lab, named directly after Bender's and her co-authors' term. It was this widespread, but often oversimplified use of the metaphor that prompted Bender to write a clarifying post for the fifth anniversary of the article.

What's wrong with the term "artificial intelligence"

One section of Bender's interview focuses on the very concept of "artificial intelligence" — in her opinion, it is just as misleading as the oversimplified interpretations of "stochastic parrot." She describes herself as a researcher studying how language works and how people use it; in computational linguistics, she teaches students who will go on to create language technologies.

"Linguistics in the broadest sense is the study of how language works and how we work with language.

I contribute to this, and I also work in computational linguistics, teaching students who will create language technologies," said Emily Bender in an interview with IEEE Spectrum.

This framing of the task is important for understanding her original critique: Bender speaks not as an opponent of technology in general, but as a linguist insisting on the precision of language used to describe the capabilities of language models — and it is precisely this lack of precision in language, in her opinion, that underlies most misconceptions about the 2021 article.

What this means

Five years after publication, the "stochastic parrot" metaphor lives its own life in popular AI discourse, having diverged from its original academic context. Through her post and interview, Bender is trying to restore precision of language to the conversation, precision that was the foundation of her original critique of large language models.

Frequently Asked Questions

What does the term "stochastic parrot" mean?

It's a metaphor from the 2021 article: large language models statistically predict probable sequences of words, but do not understand the meaning of what is said — that is, they "repeat" language like a parrot, rather than comprehend it.

Why is the "stochastic parrots" article called controversial?

Because shortly before its publication in March 2021, Google fired two of the four authors — Timnit Gebru and Margaret Mitchell — which attracted additional attention from the media and academic community to the work.

Where did

Emily Bender explain misconceptions about the article?

She wrote the post "Stochastic Parrots: Frequently Unasked Questions" and separately discussed her arguments in an interview with IEEE Spectrum, timed to the fifth anniversary of the original article's publication.