On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?
Bender et al., 2021, On the Dangers of Stochastic Parrots (Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (FAccT ’21))
Subscriber deck — free during early access
1 cards in “4.2 Static Data/Changing Social Views” · whole deck
4.2 Static Data/Changing Social Views
- What is “value-lock” in Bender et al.?Social movements create new language and reframe narratives (e.g. Black Lives Matter influenced Wikipedia coverage of police shootings), but LMs trained on static data reify older, less inclusive understandings, and retraining huge models often enough to keep up is infeasible.
- Bender et al., 2021, On the Dangers of Stochastic Parrots (Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (FAccT ’21)), §4.2, p. 614
Source
- Emily M. Bender, Timnit Gebru, Angelina McMillan-Major and Shmargaret Shmitchell. 2021. On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?. In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (FAccT ’21), pp. 610–623. License: CC BY 4.0. Flashcards adapted from this work. doi:10.1145/3442188.3445922