Emily M. Bender, Professor, University of Washington
5.0/10
Emily M. Bender, Professor, University of Washington
Emily M. Bender is an American linguist who works on multilingual grammar engineering, technology for endangered language documentation, computational semantics, and methodologies for supporting consideration of impacts language technology in NLP research, development, and education. She is the Howard and Frances Nostrand Endowed Professor of Linguistics at the University of Washington. Her work includes the LinGO Grammar Matrix, an open-source starter kit for the development of broad-coverage precision HPSG grammars.
Since 2003, Dr. Bender has served as an instructor at the University of Washington. She is presently the faculty director of the CLMS program, the head of the Computational Linguistics Laboratory, an adjunct professor in the Department of Computer Science and Engineering, and a professor in the Department of Linguistics. She worked in industry at YY Technologies and had interim posts at Stanford University and UC Berkeley before coming to UW. She joined the HPSG and LinGO programs at CSLI after receiving her PhD from Stanford University’s Department of Linguistics. She received her undergraduate degree from UC Berkeley in linguistics, and she has also studied at Tohoku University in Sendai, Japan.
2023
Co-authored with Google researcher Timnit Gebru and others, Bender presented a paper in 2021 at the ACM Conference on Fairness, Accountability, and Transparency titled “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜” that Google attempted to prevent from being published. This was a part of a series of events that culminated in Gebru’s departure from Google, the specifics of which are up for debate. In order to construct natural language processing systems employing machine learning from massive text corpora, the research addressed ethical concerns. Since then, she has worked to raise awareness of AI ethics and has resisted the hoopla surrounding massive language models.
She makes a difference between meaning and form in language. Meaning is the concepts that language conveys, whereas form is the structure of language, such as grammar. She made the case in a 2020 paper that machine learning models for natural language processing that are trained solely on form and have no relation to meaning are unable to comprehend language in a meaningful way. She has therefore maintained that programs such as ChatGPT are incapable of producing meaningfully understood text from the material they analyze.
In 2021, Bender was chosen as the Association for Computational Linguistics’ vice-president-elect. Before becoming Vice-President in 2023, President in 2024, and Past President in 2025, Bender tried hold the office of Vice President-elect in 2022. The American Association for the Advancement of Science chose Bender as a Fellow in 2022.
Reflecting on this a bit further — I think this episode really emblematic of how broken evaluation is in our field right now. In general, data work is important, and creating benchmarks is work that has been historically undervalued. However: https://t.co/BQShkoDjTN
If Google were following best practice in data and model development and documentation, the dataset documentation would have been done before training even started. A reminder, from Stochastic Parrots: https://t.co/DFsdBHTGNLpic.twitter.com/SsJZZafMyb
