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
2 cards in “6.1 Coherence in the Eye of the Beholder” · whole deck
6.1 Coherence in the Eye of the Beholder
- Why is LM-generated text only “seemingly coherent”?Human communication is jointly constructed: listeners model the speaker’s beliefs, intentions and shared ground. LM text has no communicative intent, world model or model of the reader; the coherence we perceive comes from our own linguistic competence and predisposition to find meaning.
- Bender et al., 2021, On the Dangers of Stochastic Parrots (Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (FAccT ’21)), §6.1, pp. 616–617
- What is a “stochastic parrot”?Bender et al.’s description of an LM: a system that haphazardly stitches together sequences of linguistic forms observed in its training data, according to probabilistic information about how they combine, but without any reference to meaning.
- Bender et al., 2021, On the Dangers of Stochastic Parrots (Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (FAccT ’21)), §6.1, pp. 616–617
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