In machine learning, the term stochastic parrot is a metaphor, introduced by Emily M. Bender and colleagues in a 2021 paper, that frames large language models as systems that statistically mimic text without real understanding.[1][2] The term carries a negative connotation.[2]
Origin and definition
The term was first used in the paper "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜" by Bender, Timnit Gebru, Angelina McMillan-Major, and Margaret Mitchell (using the pseudonym "Shmargaret Shmitchell").[1][2] They argued that large language models (LLMs) present dangers such as environmental and financial costs, inscrutability leading to unknown dangerous biases, and potential for deception, and that they can't understand the concepts underlying what they learn.[3]
The word "stochastic" – from the ancient Greek "στοχαστικός" (, "based on guesswork") – is a term from probability theory meaning "randomly determined".[2] The word "parrot" refers to parrots' ability to mimic human speech, without understanding its meaning.[2]
Usage
The phrase has been used by AI skeptics to signify that LLMs lack understanding of the meaning of their outputs.
Sam Altman, CEO of OpenAI, used the term shortly after the release of ChatGPT, when he tweeted "i am a stochastic parrot, and so r u".[2] The term was designated to be the 2023 AI-related Word of the Year by the American Dialect Society.[6][7]
Debate
Some LLMs, such as ChatGPT, have become capable of interacting with users in convincingly human-like conversations.[8] The development of these new systems has deepened the discussion of the extent to which LLMs understand or are simply "parroting".
Subjective experience
In the mind of a human being, words and language correspond to things one has experienced.[9] For LLMs, according to proponents of the theory, words correspond only to other words and patterns of usage fed into their training data.[10][11][1] Proponents of the idea of stochastic parrots thus conclude that statements about LLMs are due to "the human tendency to attribute meaning to text",
See also
- Chinese room
- Criticism of artificial neural networks
- Criticism of deep learning
- Generative AI
- Mark V. Shaney, an early chatbot that used a very simple three-word Markov chain algorithm to generate Markov text
- Autocomplete
Further reading
External links
- "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜" at Wikimedia Commons
References
- —^
- Ben Zimmer. 'Stochastic Parrot': A Name for AI That Sounds a Bit Less Intelligent Wall Street Journal, 2024-01-18, retrieved 2024-04-01^
- Karen Hao. We read the paper that forced Timnit Gebru out of Google. Here's what it says. MIT Technology Review, 4 December 2020, retrieved 19 January 2022^