Stochastic parrot
In machine learning, the term stochastic parrot is a metaphor that frames large language models as systems that statistically mimic text without real understanding. The word "stochastic" – from the ancient Greek "στοχαστικός" (stokhastikos, 'based on guesswork') – is a term from probability theory meaning "randomly determined". The word "parrot" refers to parrots' ability to mimic human speech.
The term was introduced in a 2021 paper on AI ethics titled "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜" that was authored by Timnit Gebru, Emily M. Bender, Angelina McMillan-Major, and Margaret Mitchell. The paper outlined possible risks associated with large language models (LLMs). In December 2020, it was the subject of a workplace dispute between Gebru (then co-leader of Google's Ethical Artificial Intelligence Team) and Google, which had requested the retraction of the paper. The incident culminated in Gebru's controversial departure from the company.
The paper was later presented at the 2021 ACM Conference on Fairness, Accountability, and Transparency, and the term "stochastic parrot" has seen widespread use in academic research concerning generative AI and LLMs. The term has been interpreted negatively as an insult towards AI.