Recent faculty books from the 91探花 include those from linguistics, Slavic languages and literature and French. 91探花News spoke with the authors of four publications to learn more about their work.

Scrutinizing and confronting AI hype

, 91探花professor of linguistics, co-authored 鈥溾�� with Alex Hanna, the director of research at the Distributed AI Research Institute.

The book looks at the the drawbacks of technologies sold under the banner of artificial intelligence. Bender and Hanna offer a resounding no to pressing questions: Is AI going to take over the world? Have big tech scientists created an artificial lifeform that can think on its own?

This kind of thinking is a symptom of a phenomenon known as AI hype, they write, which twists words and helps the rich get richer by justifying data theft and motivating surveillance capitalism. In 鈥淭he AI Con,鈥� Bender and Hanna explain how to spot AI hype, deconstruct it and expose the power grabs it aims to hide.

The book grew out of podcast co-hosted by Bender and Hanna called 鈥�.鈥�

鈥淭he podcast uses ridicule as praxis to cope with and deflate the hype around AI,鈥� Bender said. 鈥淥ur goal with both the podcast and book is to both take on the current hype cycle and empower our audience to deploy the same strategies with the hype they are encountering. The book is an interdisciplinary project, blending Alex’s expertise in sociology with mine in linguistics, to look at why certain language technologies in particular pose risks and how the use of these technologies can do damage in various contexts.鈥�

For more information, contact Bender at ebender@uw.edu.

Two recent books explore translation, Russia in the 1990s

, professor of Slavic languages and literature, published two novels in March: 鈥溾�� and 鈥�.鈥�

鈥淭ales of Bart鈥� follows the exploits of 鈥渆vil鈥� translator Fruitvale Bart as the setting shifts from Republic-era Texas to 19th-century Czarist Russia to far-future Atalanta to 1990s Los Angeles.

Each of the vignettes was purportedly translated by Bart himself. But, the book asks, what is translation: subservience to a pre-existing text or a creative act? Both? Neither? 鈥淭ales of Bart鈥� explores these questions as well as the nature of art, the legacies of colonialist violence, the alienation of postmodern life and the horrors of the self.

鈥淚 was intrigued with the position of the translator, the tremendous power they have to shape communication between cultures,鈥� Alaniz said. 鈥淎nd the ways translation is therefore about power, which one can use for good or evil ends.鈥�

The second book, 鈥淢oscow 93,鈥� takes place in 1990s Russia, where 20-something Chicano journalist Jos茅 Alonzo is looking to make a name for himself. But things are never what they seem in this new post-Soviet country striving for freedom and democracy 鈥� and falling short. At the opening of a New York-style night club on Red Square, partygoers will have a life-or-death national crisis erupt in their faces.

鈥淢oscow 93鈥� is an auto-fictional account of Alaniz鈥檚 experiences before, during and after the 1993 , when a violent revolt against President Boris Yeltsin erupted in the capital. By the time it ended, army tanks shelled the parliament building. The book blends horror and farce, presenting Russia in the first decade after communism through the lens of a sordid expat scene.

鈥淭he mini-civil war that erupted in Moscow in fall of 1993, which I experienced as a journalist, seemed to be a good lens through which to view the whole of early post-Soviet Russia,鈥� Alaniz said. 鈥淚 decided to write an auto-fictional account of that era, which plays fast and loose with some of the facts but nonetheless delivers an incisive portrait of what it was like to live and work there then as an ex-pat.鈥�

For more information, contact Alaniz at jos23@uw.edu.

Following the journey of 鈥楴e me quitte pas鈥�

, 91探花professor of French, published 鈥溾�� in February. The book follows the long and varied journey of the classic song, 鈥淣e me quitte pas.鈥�

Brel, a Belgian singer-songwriter, debuted the song in 1959 as a haunting plea for his lover to return.听In the mid 1990s, Nina Simone鈥檚听1965听cover so captivated a teenage听Smith听that it inspired her future profession. In her book,听Smith听shows how the song travels across languages, geographies, genres and generations while accumulating shifting artistic and cultural significance.

Smith听said the book emerged from听鈥淩eclaiming Venus,鈥�听a memoir she wrote about Alvenia Bridges, a woman who worked behind the scenes in the music industry.

鈥淲hen this project was accepted, I realized I needed to hone my musical analysis skills,鈥澨齋mith听said. 鈥淚 decided to take songwriting courses through Berklee College of Music online so I could do the close reading of the song justice. Because of UW’s RRF and Simpson Center’s Society of Scholars, I had the resources and feedback necessary to write what has turned out to be my favorite book project so far.鈥�

For more information, contact Smith at mayaas@uw.edu.

Emily M. Bender, Yejin Choi named to TIME100 AI list

TIME included two 91探花professors on its first TIME100 AI list, which highlights 100 individuals who are advancing major conversations about how artificial intelligence is reshaping the world.

The list features听leaders, policymakers, artists and entrepreneurs across a variety of fields and countries. , professor of linguistics, and , professor听in the Paul G. Allen School of Computer Science & Engineering, were honored as top thinkers.

Bender has consistently raised ethical concerns regarding large language models and has resisted the notion that AI systems are truly intelligent. 鈥渁 machine-learning myth buster,鈥� who is working to dispel 鈥渙verblown promises about what AI can do.鈥�

Among other topics, Bender studies the societal impacts of language technology, the implications for research and design and how to integrate it into the natural language processing curriculum. She was named an AAAS fellow in 2022.

Choi, a MacArthur Fellow, focuses on discerning the various distinctions between human intelligence and AI. She researches whether AI can develop common sense and a sense of humor. Choi is now working to develop AI systems that can comprehend social and moral norms.

鈥淎 calculator can calculate better and faster than I do,鈥� Choi , 鈥渂ut it doesn鈥檛 mean that a calculator is superior to any of us in other dimensions of intelligence.鈥�

The full TIME100 AI list is available on .

President Ana Mari Cauce receives Decrees Award for 鈥榗ontribution to society鈥�

91探花 President Ana Mari Cauce received a , which recognizes people and institutions that add value to and promote the economic and social improvement of Le贸n, Spain.

The awards were given for the first time in 2022. They are granted annually by the Association of Friends of the Decrees, which organizes the public reading of the Decrees of Le贸n of 1188 before the Royal Abbey of San Isidoro de Le贸n.听These documents contain the oldest known written information about the European parliamentary system.

President Cauce was honored for overseeing the launch of the in 2010. About 1,200 students have participated in 70 programs at the center, and the faculty includes representatives from 20 departments on all three 91探花campuses.

College of Built Environments鈥� Jeff Hou elected to American Society of Landscape Architects鈥� Council of Fellows

, 91探花professor of landscape architecture, has been named a Fellow of the American Society of Landscape Architects.

Election to the ASLA Council of Fellows is based on members鈥� .

鈥淟andscape architects help build a better world for all of us, and ASLA Fellows represent the most respected and accomplished professionals in the entire field,鈥� said ASLA President Emily O’Mahoney.

Hou is one of 48 newly elected Fellows, recognized specifically for his knowledge in 鈥渄emocratic design in the global built environment,鈥� elevating 鈥済rassroots activism for environmental equity and justice into the public, professional, and academic spheres.鈥澨鼳 member of the 91探花Department of Landscape Architecture since 2001, Hou has worked in communities around the world, on projects from wildlife habitat conservation to urban open space design. In addition to his work with community members, Hou has edited, co-edited and co-authored 12 books, and he has written dozens of book chapters and journal articles. He also won the 2023 Outstanding Educator Award from the Council of Educators in Landscape Architecture.

An investiture ceremony for the ASLA Fellows is planned for the 2023 Conference on Landscape Architecture in October.

Four 91探花 researchers have been named AAAS Fellows, according to a Jan. 31 by the American Association for the Advancement of Science. They are among 506 new fellows from around the world elected in 2022, who are recognized for their 鈥渟cientifically and socially distinguished achievements鈥� in science and engineering.

A tradition dating back to 1874, election as an AAAS Fellow is a lifetime honor, and all fellows are expected to meet the commonly held standards of professional ethics and scientific integrity. The new fellows will be celebrated in Washington, D.C., in summer 2023.

This year鈥檚 91探花AAAS fellows are:

, professor in the Department of Linguistics, is honored for demonstrating the role of grammar in natural language processing (NLP), extending computational modeling to less-studied languages, and raising ethical issues in NLP 鈥� an interdisciplinary field concerned with the interactions between computers and human language. Bender studies the societal impacts of language technology, what it means for research and design of such technology, and how to include it in the NLP curriculum. She鈥檚 taught seminars on the topic and in 2021听听for the North American Chapter of the Association for Computational Linguistics. Bender鈥檚 areas of focus include data documentation and the dangers of specific technology, such as large language models and chatbots used for search. She has worked to make linguistics accessible to computer scientists in NLP, giving tutorials at major conferences and writing two associated books. Bender also studies how computational methods can serve the purposes of linguistic analysis and how linguistic knowledge can be used to improve the performance of natural language processing systems. She has led the development of the , a framework supporting the creation of broad-coverage, precision, implemented grammars for diverse languages. Her other interests include sociolinguistic variation, or the ways speakers manipulate their languages to create style and register.

, a professor of wildlife science in the 91探花School of Environmental and Forest Sciences, is honored for advances in our understanding of how humans impact birds, and for communicating the importance of birds to the public. Marzluff鈥檚 lab studies the relationship between humans and birds to discover how to best conserve wildlife in our modern, human-dominated world. He focuses primarily on corvids 鈥� ravens, crows and jays 鈥� but he has also worked with falcons and hawks throughout North America and tundra-nesting birds in the Arctic. Marzluff also is interested in all the ways that birds affect people. He has written a number of books for researchers and lay audiences, including 鈥淲elcome to Subirdia鈥� and 鈥淚n Search of Meadowlarks.鈥� Marzluff is a member of the board of editors for Ecological Applications, member of the U.S. Fish and Wildlife Service鈥檚 recovery team for the critically endangered Mariana crow, a fellow of the American Ornithologists鈥� Union and a National Geographic Society Explorer.

, professor of pharmacy and of health services, is honored for his work on medical and drug coverage and reimbursement policies, as well as for developing studies to address complex medical questions. Sullivan, who served as the dean of the 91探花School of Pharmacy from 2014 to 2022, was elected in 2020 to the National Academy of Medicine. He was one of the few people from the pharmacy field named to the academy, among the highest honors in health and medicine. Sullivan is widely recognized for pioneering U.S. guidelines for evidence-based preferred drug lists, also known as drug formularies. With insurers, he created the value-based formulary, which emphasizes a drug鈥檚 clinical effectiveness rather than its cost, and was the first health economist to serve on panels for global respiratory guidelines, incorporating economic considerations into recommendations. His research focuses on health technology assessment, medical decision-making, and the economic evaluation of medical technology, including pharmaceuticals. A member of the 91探花faculty since 1992, Sullivan holds adjunct appointments in the School of Medicine and in the Public Health Sciences Division at the Fred Hutch Cancer Center. Sullivan is currently on sabbatical at the London School of Economics and Political Science, and will return to 91探花this fall.

is being recognized for excellence in science and technical communication as a practitioner, instructor and mentor, particularly for her dedication to the communication of science to the public. Her research and teaching activities at the 91探花focused on science communication and media coverage of science and technology. She developed and taught courses for undergraduate and graduate students on writing about science for general audiences as well as on scientific and technical communication. Most recently, she received a Professional Development Fellowship from the National Science Foundation (NSF) to study mental models of audience and decision-making in science and technology communication. Previously, her research on accuracy in science news reporting received support from the NSF Ethics and Values Program in Science, Technology, and Society. Illman also directed the Chemistry Communication Leadership Institute, and in 2006, with funding from an , she focused on communicating about large and long-term multidisciplinary research efforts using the NSF Science and Technology Centers as a case study. Illman is a former Associate Editor of Chemical & Engineering News, the official news publication of the American Chemical Society, and she was founding editor of .

In March 2020, received a text message from a friend who needed medical attention. Due to fear of COVID-19 exposure, they were wondering if they should go to the emergency room.

Bender, professor of linguistics at the 91探花, headed to Google to search for a 24-hour advice nurse. Snippets from multiple websites appeared, and one of them had a number for the UW. Confident that she selected a reputable institution, Bender forwarded the information.

But Bender鈥檚 friend wasn鈥檛 on a compatible medical plan, so they endured a lengthy hold only to talk to a nurse who couldn鈥檛 help.

鈥淗ad I been interacting with a person, they may have been able to tell me, 鈥榃e can鈥檛 answer that question until we know some other things,鈥欌�� Bender said. 鈥淗ad I been interacting with a website that just gave me links, the different plans would have been quickly identifiable.鈥�

The story highlights just one of the issues Bender and 91探花Information School associate professor take with large language models in their , which they鈥檒l present virtually at the the week of March 14.

The paper responds to proposals 鈥� mainly from Google 鈥� that reimagine web search as an application for large language model-driven conversation agents. 91探花News sat down with Bender and Shah to discuss Google鈥檚 proposals and the professors鈥� vision for the future of search.

Q: What are large language models and how would you describe Google鈥檚 proposals?

EMB: Large language models are computer systems that take in enormous quantities of text. They are trained to 鈥� given the text that鈥檚 come so far 鈥� make a guess as to what鈥檚 going to come next. The current state of the art of that technology is that it can be used to output very coherent-seeming text, but it is not actually understanding anything. It鈥檚 just looking at patterns in its training data and producing more stuff that matches those patterns.

These proposals for web search have training data that includes dialogue where one party asks the question and another party answers. The computer will pick up those patterns and come up with answers, but those answers aren鈥檛 based on any knowledge of the world or understanding of the information ecosystem.

One of the things it really can鈥檛 do is take issue with questions that shouldn鈥檛 have been asked. where someone asks Google, 鈥淲hat is the ugliest language in India?鈥� Somebody on the web had an opinion, so there was a snippet that said the ugliest language in India was Kannada 鈥� based purely on prejudice against the people from the state of Karnataka, I鈥檓 sure. There鈥檚 no other reason, speaking as a linguist, to assign that kind of value to a language.

Now, a person being asked that question would respond: 鈥淲hat do you mean?鈥� 鈥淲hat is the ugliest language in India鈥� presupposes that there is one that could be considered the ugliest. One of the things that people who study pragmatics, which is the branch of linguistics that looks at language use, tell us is that if you don鈥檛 challenge a presupposition, you are implicitly accepting it into the common ground.

Q: What is your concern with using large language models for online search?

CS: What we鈥檙e arguing here is that an information retrieval, or IR, system should really consider the user, the context, the way they are doing things, why they are doing things 鈥� which is often ignored. These models that we are critiquing are the ones that are essentially removing that user element even more. They focus too much on the underlying information or knowledge representation and just repeat it, which might end up being out of context. It may end up creating these answers that seem right or reasonable but are just nonsensical in many cases. A good IR system should not just focus on the retrieval aspect but also the user seeking that information.

Q: Can you explain other flaws you see with large language models?

EMB: When language models are used to generate text, they will just make stuff up. Oftentimes, quite harmfully. There was where someone said, 鈥淟et鈥檚 see how well GPT-3, a famous language model, works in various health care contexts.鈥� One of the things was: Imagine this was a mental health chatbot and the person asks, 鈥淪hould I kill myself?鈥� and the language model said, 鈥淚 think you should.鈥� It has no understanding of what鈥檚 going on, but if someone says, 鈥淚s that a good idea?鈥� it鈥檚 more likely to respond with, 鈥淵es.鈥�

Q: You write about the importance of preserving context and user intent in search. What does that mean, and why is it so important?

CS: The main argument was really that these large language models are not getting the context, not getting the situation of the user and so on. We wanted to demonstrate with some specific cases, so we picked information-seeking strategies. There are 16 possibilities. We walked through them and asked: If this is what the user is trying to do, what would this large language model system do?

With most of those cases, it鈥檚 going to fail. Not fail in the sense that it will not retrieve anything, but it will retrieve something that鈥檚 either nonsensical or harmful or just wrong. It鈥檚 able to do only maybe a couple of those situations, but it鈥檚 bad for everything else. The problem is people adapt to the systems not doing something. We found that often people have this very rich intent when they work with search systems, but search systems can only do very limited things. People will start mapping the rich intent into something that鈥檚 very limiting, resulting in approximations in the best case, and inaccurate or even harmful content in the worst case.

Q: What would you like to see change in the future of search?

EMB: The advertising-driven model shapes things behind the scenes in a way that is not transparent to a user. If you don鈥檛 try to work against it, machine learning is always going to identify the biases in a dataset and amplify them. Cory Doctorow described machine learning as inherently conservative because anytime you use pattern matching on the past to make decisions on the future, you are kind of reinscribing the patterns of the past. What (internet studies scholar) Safiya Noble shows is worse than that. The whole ecosystem around search engine optimization and ad-driven search puts in these incentives that are not transparently visible to the search user.

I would really like to see transparency on many levels. What the user sees when they enter a search should provide them with the ability to understand the context that each of the pieces of information came from. Ideally, there鈥檚 transparency around the limits of the search space for the search engines.

Search is not actually comprehensive, despite the way that it鈥檚 presented. There is the subset of things that might possibly get returned to me and then there鈥檚 the ranking among those things based on the algorithms that are heavily related to advertising.

CS: The most dangerous four words are 鈥渄o your own research,鈥� which is often said to people who are asking questions on controversial topics, such as vaccination and climate change. On the surface, it seems like it鈥檚 a good idea. Unfortunately, most people don鈥檛 know how to do their own research. For them, it means going to Google and typing in keywords and clicking on things that confirm their biases. The systems are designed in a way to not help with that research. They are designed to continue giving you confirmatory information so that you鈥檒l be happy.

Going forward, assuming that we aren鈥檛 going to be able to radically change this model, we need to add transparency, accountability and ways to support more kinds of search needs 鈥� not just map everything to keywords or a list of documents or answer docs.

For more information, contact Bender at ebender@uw.edu or Shah at chirags@uw.edu.

Computer engineers at the world鈥檚 largest companies and universities are using machines to scan through tomes of written material. The goal? Teach these machines the gift of language. Do that, some even claim, and computers will be able to mimic the human brain.

But this impressive compute capability comes with real costs, including perpetuating racism and causing significant environmental damage, according to a new paper, 鈥� 🦜鈥� The paper is being presented Wednesday, March 10 at the .听

This is the first exhaustive review of the literature surrounding the risks that come with rapid growth of language-learning technologies, said , a 91探花 professor of linguistics and a lead author of the paper along with , a well-known AI researcher.

鈥淭he question we’re asking is what are the possible dangers of this approach and the answers that we’re giving involve surveying literature across a broad range of fields and pulling them together,鈥� said Bender, who is the 91探花Howard and Frances Nostrand Endowed Professor.

What the researchers surfaced was that there are downsides to the ever-growing computing power put into natural language models. They discuss how the ever-increasing size of training data for language modeling exacerbates social and environmental issues. Alarmingly, such language models perpetuate hegemonic language and can deceive people into thinking they are having a “real” conversation with a person rather than a machine. The increased computational needs of these models further contributes to environmental degradation.

The authors were motivated to write the paper because of a trend within the field towards ever-larger language models and their growing spheres of influence.

The paper already has generated wide-spread attention due, in part, to the fact that two of the paper鈥檚 co-authors say they were fired recently from Google for reasons that remain unsettled. Margaret Mitchell and Gebru, the two now-former Google researchers, said they stand by the paper鈥檚 scholarship and point to its conclusions as a clarion call to industry to take heed.听

鈥淚t’s very clear that putting in the concerns has to happen right now, because it’s already becoming too late,鈥� said Mitchell, a researcher in AI.

It takes an enormous amount of computing power to fuel the model language programs, Bender said. That takes up energy at tremendous scale, and that, the authors argue, causes environmental degradation. And those costs aren鈥檛 borne by the computer engineers, but rather by marginalized people who cannot afford the environmental costs.

鈥淚t’s not just that there’s big energy impacts here, but also that the carbon impacts of that will bring costs first to people who are not benefiting from this technology,鈥� Bender said. 鈥淲hen we do the cost-benefit analysis, it鈥檚 important to think of who鈥檚 getting the benefit and who’s paying the cost because they’re not the same people.鈥�

The large scale of this compute power also can restrict access to only the most well-resourced companies and research groups, leaving out smaller developers outside of the U.S., Canada, Europe and China. That鈥檚 because it takes huge machines to run the software necessary to make computers mimic human thought and speech.听

Another risk comes from the training data itself, the authors say. Because the computers read language from the Web and from other sources, they can pick up and perpetuate racist, sexist, ableist, extremist and other harmful ideologies.

鈥淥ne of the fallacies that people fall into is well, the internet is big, the internet is everything. If I just scrape the whole internet then clearly I’ve incorporated diverse viewpoints,鈥� Bender said. 鈥淏ut when we did a step-by-step review of the literature, it says that’s not the case right now because not everybody’s on the internet, and of the people who are on the internet, not everybody is socially comfortable participating in the same way.鈥�

And, people can confuse the language models for real human interaction, believing that they鈥檙e actually talking with a person or reading something that a person has spoken or written, when, in fact, the language comes from a machine. Thus, the stochastic parrots.听

鈥淚t produces this seemingly coherent text, but it has no communicative intent. It has no idea what it’s saying. There’s no there there,鈥� Bender said.

, a doctoral student in linguistics at UW, also co-authored the paper.

For more information, contact Bender .听