[Ads-l] Phrase: stochastic parrot. Coined by linguist Emily M. Bender

[ADSGarson O'Toole]

My candidate for the most interesting phrase of 2021 is “stochastic
parrot”. (I know this is a year too late.) If you are currently
studying computational linguistics you have probably heard this
phrase. Here is a link to the Quote Investigator article.

https://quoteinvestigator.com/2022/10/07/word-parrot/

[Begin excerpt from QI article]
Artificial intelligence (AI) researchers have used vast amounts of
text to train digital neural networks which capture the intricate
statistical patterns of word sequences. The resultant systems are
called large language models. One of the most famous is GPT-3
(Generative Pre-trained Transformer 3).

Language models (LMs) are able to perform a variety of tasks, e.g.,
answering questions, summarizing documents, generating text, and
translating text. However, influential AI researchers believe these
capabilities are misleading and often overestimated. Thus, these
models should be considered merely “stochastic parrots”.
. . .
In March 2021 Emily M. Bender, Timnit Gebru, Angelina McMillan-Major,
and Margaret Mitchell (who used the pseudonym Shmargaret Shmitchell)
published “On the Dangers of Stochastic Parrots: Can Language Models
Be Too Big?” which critically examined recent research efforts. The
title included a parrot emoji:

[Begin nested quotation]
Contrary to how it may seem when we observe its output, an LM is a
system for haphazardly stitching together sequences of linguistic
forms it has observed in its vast training data, according to
probabilistic information about how they combine, but without any
reference to meaning: a stochastic parrot.
[End nested quotation]
[End excerpt]

I contacted Emily M. Bender via twitter, and she told me that she
coined the memorable phrase “stochastic parrot”.

Garson

------------------------------------------------------------
The American Dialect Society - http://www.americandialect.org