Stochastic parrot – Wikipedia
From Wikipedia, the free encyclopedia
Time period utilized in machine studying
In machine learning, “stochastic parrot” is a time period coined by Emily M. Bender[2][3] within the 2021 artificial intelligence analysis paper “On the Risks of Stochastic Parrots: Can Language Fashions Be Too Massive?” by Bender, Timnit Gebru, Angelina McMillan-Main, and Margaret Mitchell.[4] The time period refers to “giant language fashions which can be spectacular of their means to generate realistic-sounding language however in the end don’t really perceive the that means of the language they’re processing.”[2]
Definition and implications[edit]
Stochastic means “(1) random and (2) involving likelihood or chance”.[5] A “stochastic parrot”, in accordance with Bender, is an entity “for haphazardly stitching collectively sequences of linguistic kinds … in accordance with probabilistic details about how they mix, however with none reference to that means.”[3] Extra formally, the time period refers to “giant language fashions which can be spectacular of their means to generate realistic-sounding language however in the end don’t really perceive the that means of the language they’re processing.”[2]
In accordance with Wahlström, et. al., the analogy highlights two very important limitations:
(i) The predictions made by a studying machine are primarily repeating again the contents of the info, with some added noise (or stochasticity) attributable to the restrictions of the mannequin.
(ii) The machine studying algorithm doesn’t perceive the issue it has learnt. It may’t know when it’s repeating one thing incorrect, out of context, or socially inappropriate.
They go on to notice that due to these limitations, a studying machine would possibly produce outcomes that are “dangerously unsuitable”.
The time period was first used within the paper “On the Risks of Stochastic Parrots: Can Language Fashions Be Too Massive?” by Bender, Timnit Gebru, Angelina McMillan-Main, and Margaret Mitchell (utilizing the pseudonym “Shmargaret Shmitchell”).[4] The paper coated the dangers of very large language models, relating to their environmental and monetary prices, inscrutability resulting in unknown harmful biases, the shortcoming of the fashions to know the ideas underlying what they be taught, and the potential for utilizing them to deceive folks.[6] The paper and subsequent occasions resulted in Gebru and Mitchell losing their jobs at Google, and a subsequent protest by Google workers.[7][8]
Subsequent utilization[edit]
In July 2021, the Alan Turing Institute hosted a keynote and panel dialogue on the paper. As of Might 2023[update], the paper has been cited in 1529 publications.[10] The time period has been utilized in publications within the fields of legislation,[11] grammar,[12] narrative,[13] and humanities.[14] The authors proceed to keep up their issues concerning the risks of chatbots primarily based on giant language fashions, resembling GPT-4.[15]
See additionally[edit]
References[edit]
- ^ a b c Uddin, Muhammad Saad (April 20, 2023). “Stochastic Parrots: A Novel Look at Large Language Models and Their Limitations”. In direction of AI. Retrieved 2023-05-12.
- ^ a b Weil, Elizabeth (March 1, 2023). “You Are Not a Parrot”. New York. Retrieved 2023-05-12.
- ^ a b Bender, Emily M.; Gebru, Timnit; McMillan-Main, Angelina; Shmitchell, Shmargaret (2021-03-01). “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?”. Proceedings of the 2021 ACM Convention on Equity, Accountability, and Transparency. FAccT ’21. New York, NY, USA: Affiliation for Computing Equipment: 610–623. doi:10.1145/3442188.3445922. ISBN 978-1-4503-8309-7. S2CID 232040593.
- ^ “Stochastic”. Merriam-Webster. Retrieved 2023-05-13.
- ^ Haoarchive, Karen (4 December 2020). “We read the paper that forced Timnit Gebru out of Google. Here’s what it says”. MIT Technology Review. Archived from the unique on 6 October 2021. Retrieved 19 January 2022.
- ^ Lyons, Kim (5 December 2020). “Timnit Gebru’s actual paper may explain why Google ejected her”. The Verge.
- ^ Taylor, Paul (2021-02-12). “Stochastic Parrots”. London Review of Books. Retrieved 2023-05-09.
- ^ “Bender: On the Dangers of Stochastic Parrots”. Google Scholar. Retrieved 2023-05-12.
- ^ Arnaudo, Luca (April 20, 2023). “Synthetic Intelligence, Capabilities, Liabilities: Interactions within the Shadows of Regulation, Antitrust – And Household Legislation”. SSRN. doi:10.2139/ssrn.4424363. S2CID 258636427.
- ^ Bleackley, Pete; BLOOM (2023). “In the Cage with the Stochastic Parrot”. Speculative Grammarian. CXCII (3). Retrieved 2023-05-13.
- ^ Gáti, Daniella (2023). “Theorizing Mathematical Narrative via Machine Studying”. Journal of Narrative Theory. Venture MUSE. 53 (1): 139–165. doi:10.1353/jnt.2023.0003. S2CID 257207529.
- ^ Rees, Tobias (2022). “Non-Human Phrases: On GPT-3 as a Philosophical Laboratory”. Daedalus. 151 (2): 168–82. doi:10.1162/daed_a_01908. JSTOR 48662034. S2CID 248377889.
- ^ Goldman, Sharon (March 20, 2023). “With GPT-4, dangers of ‘Stochastic Parrots’ remain, say researchers. No wonder OpenAI CEO is a ‘bit scared’“. VentureBeat. Retrieved 2023-05-09.
Works cited[edit]
Additional studying[edit]
- Thompson, E. (2022). Escape from Mannequin Land: How Mathematical Fashions Can Lead Us Astray and What We Can Do about It. Primary Books. ISBN 978-1541600980.
Exterior hyperlinks[edit]
