Maria Antoniak

@mariaa.bsky.social

11674 Followers 1419 Following

NLP, cultural analytics. Assistant Professor of CS at University of Colorado Boulder. Books, bikes, games, art. https://maria-antoniak.github.io

Screenshot of a Bluesky account, “Argos News” aka render87.bsky.social with a bunch of labels showing that it is posting nonstop without breaks. The bio reads “Global news bot, powered by AI.”
Labels on the “render87.bsky.social” account showing that the account is posting nonstop without breaks.
A set of disallowed words and hashtags including "neuro linguistic" and "#neurolinguisticprogramming".
List of URLs
List of exact word matches.
YOU ARE... The Kontextmaschine patron saint: kontextmaschine It's real, it's powerful, and you find the panic about it as tedious as the hype. You're not interested in whether AI is good or bad — you're interested in who captures the value, who loses their leverage, and which institutions are using it as a story to do what they were already doing. Same as it ever was.
See the main page of the linked web explorer.
Figure 1: A web passage scored across our 12 narrative dimensions. Agency and setting dimensions are rated on a 1–5 Likert scale, temporal sequencing and causal density are passage-level proportions (0–1), and event density is the rate of event triggers per token. This passage scores high on agency and event features but low on setting, a “narrative profile” commonly seen across first-person web narratives.
A screenshot of my results. “MARIA ANTONIAK NLP AND COMPUTATIONAL SOCIAL SCIENCE RESEARCHER 537 STRENGTH · TOP 8% GPT-5.5 SAYS COMPUTER SCIENTIST AND NLP RESEARCHER KNOWN FOR WORK ON COMPUTATIONAL SOCIAL SCIENCE AND NATURAL LANGUAGE PROCESSING”
A fluffy black cat looks at the camera. It’s sitting on a wood floor surrounded by full bookshelves.
Please note the new ARR originality requirement and overlap with other work by the same authors (also can be found in the ARR Call for Papers): “All submitted papers are expected to describe original, completed and unpublished work. See the publication ethics guidelines on re-use of text from other publications (no more than 10% total tokens). There can be no overlap in stated contributions. ARR now follows ICML in discouraging thinly sliced contributions as follows: any concurrently submitted papers on a related topic with an overlapping set of authors must cite each other and discuss the differences in the related work section. Anonymized versions of such papers should be uploaded as supplementary material. Recommended citation format in the bibliography: Anonymous (2026). Paper title. Under review. Papers violating the above originality requirements may be desk rejected.” In addition, all authors involved may be ineligible to commit their paper(s) to EMNLP 2026 and EMNLP 2027.
The cover of the fourth book in the series. It again is an abstract "heat map" of blues, yellows, and reds, with the heat center inside an oval containing the title and author.
System: You are an expert Named Entity Recognition (NER) system. Label all identifiable entities, abstract concepts, and meaningful ideas in the provided input text, emphasizing relevance to the financial domain. Ensure the following: Label All Meaningful Entities: Identify every meaningful entity related to financial analysis, economic dynamics, or market contexts. Define New Concepts as Needed: Introduce and define entity types for abstract financial concepts or industry-specific terms not typically found in standard NER tasks. Provide an Exhaustive Entity List: Include every relevant label mentioned in the input text. Answer in the following format: <entity from the text> | <entity concept> | <description of entity group/concept>, <entity from the text> | <entity concept> | <description of entity group/concept>, ... Here is an Example : Input: Lawmakers continue to try to police social media use among teens — but Meta, parent company to Facebook, Instagram, and Threads, is ...
It happened one day that an atheist and a man were standing together on a doorstep; and the atheist said, “It is raining.” To which the man replied, “What is raining?” Now the same anarchic mystery that clings round the phrase, “il pleut,” clings round the phrase, “il faut.” In English it is generally represented by the passive mood in grammar, and the Eugenists and their like deal especially in it; they are as passive in their statements as they are active in their experiments. Lady Macbeth said, “Give me the daggers; I’ll let his bowels out.” The Eugenist would say, “In such cases the bowels should, etc.” But when the Eugenists say, “Conditions must be altered” or “Ancestry should be investigated,” or what not, it seems clear that they do not mean that the democracy must do it, whatever else they may mean. The question remains, therefore, whom they do instinctively trust when they say that this or that ought to be done. What is this flying and evanescent authority that vanishes wherever we seek to fix it? Who is the man who is the lost subject that governs the Eugenist’s verb?
A screenshot of Pagebound showing a navigation bar with search, library, discover, and discuss. underneath are two cards, each featuring some reading activity from another user. The color scheme is pastel and the style is very cute.
The cover of the book. It shows a misty forest and large stone tower.
A famous meme photo of a raccoon with a grin and tapping its fingers together, as if it is plotting something.
Screenshots of Eva Stratt (a character in Project Hail Mary) wearing different outfits and making different facial expressions.
A screenshot of Lea showing a moderation panel where users can block other users based on who liked or reposted a Bluesky post. The tool warns the user that this list includes people that they follow and that follow them and checks for permission.
ChatGPT Edu is now available to eligible CU Boulder faculty and staff. This optional, higher‑education–focused generative AI platform is designed to protect academic freedom and privacy, provide equitable access, and offer a more secure alternative to publicly available AI tools. Student access has been delayed until August 14, following input from faculty governance, as announced in a message from the university president and campus chancellors. A platform designed for higher education ChatGPT Edu is provided through a University of Colorado systemwide agreement with OpenAI, ensuring access to a FERPA‑compliant, privacy‑aware environment designed specifically for higher education. Information entered into CU Boulder’s ChatGPT Edu environment is not used to train AI models, helping reduce data security risks associated with consumer tools and better protect student, faculty, staff and university data. Already using ChatGPT? Make the switch Generative AI tools are already widely used across campus. Before the OpenAI agreement was signed, more than 28,000 students, faculty and staff created ChatGPT accounts using CU Boulder email addresses, with many more likely using personal accounts. ChatGPT Edu offers a safer, institutionally supported alternative than the publicly available version, providing stronger privacy protections while still delivering the capabilities users expect. CU Boulder encourages faculty and staff currently using the public version of ChatGPT to consider switching to ChatGPT Edu to benefit from its more data‑secure environment.
We have a commercial relationship with Piazza Technologies Inc. (“Piazza”). Pooja Sankar is the Chief Executive Officer of Piazza and is the spouse of Shyam Sankar, one of our executive officers. During 2017, 2018, 2019, and the six months ended June 30, 2019, we made payments of $412,000, $202,000, $167,000, and $135,000, respectively, for a license to its online recruiting platform and other services provided by Piazza.
Erin Kissane speaking on the Streamplace live stream. Her slides show an illustration of kelp (by her daughter!) and text saying "Shelter sensemaking: We could build sheltered spaces for community knowledge-making (which would also make a lot of the big-world hostility less necessary)"
claude prompt asking for prettier colors
cover of volume 1: each cover shows an abstract shape with swirling colors
Hope Springs Eternal • My colleagues and I always hoped that linguistics will eventually allow us to strike gold – HMM tagging (Bahl & Mercer 1976) • The quote accentuated a certain situation that existed in ASR in the seventies and in NLP in the eighties
We evaluate papers using various procedures: Moderator assessment of minimal contribution to social science Identification of author as a human with a consistent identity (as marked by ORCiD or other sources) Moderator detection of likely-AI slop, with the following red flags: Literature reviews or synthesis without substantive contribution Unsubstantiated grand theories, which may have: Fanciful equations Integration of multiple major theories into definitive new “paradigms,” “proposed frameworks,” or similar, without substantive engagement Papers written in many short sections, often including bullet-point lists and numbered paragraphs. Fake empirical research Assessment, whether by machine or by moderator perception, of papers with: High truth-indifference or phoniness (e.g., no realistic grounding) Enterprise misrepresentation (e.g., false or misleading description of the research project) Constraint evasion (e.g., implausible or facile assumptions without evidence) Low falsifiability (e.g., untestable claims without justification) Absence of methodological accountability (e.g., extensive methods and data discussion without any research output)
Screenshots of Claude’s “thinking” while it loads its responses. For example: “Vibing…” “Spelunking…”
In early 2024, researchers were already heavily using AI for work - Survey of 816 verified authors via Semantic Scholar - 81% of researchers reported using LLMs in their workflow - Top uses: information seeking & editing - Rare for data tasks: 6973% never use LLMs for data cleaning or generation
A blizzard warning has been issued for all five boroughs of New York City, Long Island and much of NJ as a classic nor'easter threatens to dump heavy snow and strong winds on our area. A blizzard warning is issued when snow accumulations are expected between 13 and 18 inches and winds are expected to gust as high as 55 mph. Whiteout conditions are expected. Forecasts now show up to a foot of snow for the five boroughs, along with central Jersey and Nassau County. Some farther east may get more than a foot of snow. Totals may be a little lower farther north and west. The blizzard warning takes effect at 6 a.m. Sunday and lasts through 6 p.m. Monday.
Weather predictions for the coming week in Boulder, Colorado. Temperatures are in the 60s and 70s with sunshine every day.
photo from article showing pasta cooked with different methods (more water, more salt, etc)
We on the left generally agree that education is for the student’s benefit, not for the benefit of their future employer, and that students go to school not merely to acquire skills but to develop an entire social and intellectual life: to have something good and to have it forever. We are sometimes embarrassed to say this, I think, out of misplaced or excessive courtesy: we have seen too many snobs tell people what they ought to like. But we shouldn’t be. It is not snobbish to say that a person with lungs must breathe or that a person with a stomach must eat, nor that a person with a mind must think. It is not snobbish to show someone how to love something new—it is a gift.
A screenshot of the remastered Horizon Zero Dawn game, showing a woman with long red hair riding a robot animal in an alpine valley.
Section 3.1 of the paper.
A screenshot of a website pane showing trending paper discussions. The first two papers are "The Issue with Special Issues..." and "Paper Skygest: Personalized..."
A moderation panel with alerts showing that my post is getting a lot of interaction and that a high profile account reposted my post.
Screenshot of a Bluesky post that reads “what’s stopping russia from doing a targeted operation to go steal zelensky in a few weeks now?”
A screenshot of a list of followers but all the photos and names and bios are replaced by "Invalid Handle."
The title slide, "Poetry as Test, Poetry as Text" by Maria Antoniak at the University of Colorado Boulder.
A view looking down at Copenhagen, with fancy bicycle parking in the foreground.
Painting by Georgia O’Keefe featuring swirling dark colors.
accounts that i mentioned the most this year, including Graze, Kagi, and Shaily Bhatt
word cloud showing my top words for the year, including models, paper, hard, research
The rating and title of an Amazon book review. The user assigned four stars and titled the review “Too much poetry. Too many characters.”
A screenshot of Catherine Yeh's website: https://catherinesyeh.github.io/ The website looks clean, colorful, and modern.
Title, authors, and abstract of paper
Title, authors, and abstract of the paper
Title, authors, and abstract of the paper.
A slide that reads “NLP and all the interesting and weird ways it intersects with processes and values that comprise democracy”
A photo of a pathway in autumn trees.
Table 2 from the linked paper, showing the set of sensitive topics and their distribution.
Figure 1 from the paper. "Distribution of the CD5 index with vs without the hidden outliers and its impact on the apparent decline of disruptive science and technology. This figure shows that CD5 = 1 papers and patents are driving the reported decline in the disruptiveness of scientific and technological knowledge over time for the Web of Science data source (with 22, 479, 429 papers) and the PatentsView data source (with 2, 926, 923 patents). For PatentsView, we also have access to sufficient metadata to exclude patents that make zero references, similarly impacting the decline. a, The distribution of the CD5 index for papers in Web of Science as presented in Park et al. [1], created using the binwidth parameter in seaborn 0.11.2. This version of the library contains a bug regarding silently dropping the largest data points (1 in this case) when specifying the binwidth parameter [3]. b, The correct histogram for papers when using the bins parameter in seaborn 0.11.2. A peak at CD5 = 1 is revealed with 972, 161 additional papers. c, The time evolution of the average CD5 index for papers. When dropping the hidden outliers with CD5 = 1, the decline in disruptiveness almost completely disappears. The shaded bands correspond to 95% confidence intervals. Finally, note that the curve without CD5 = 1 papers corresponds to (a), the histogram presented in Park et al. [1]. d–f, The equivalent plots for PatentsView revealing 142, 362 additional patents with CD5 = 1. When dropping the outliers with CD5 = 1, the decline in disruptiveness reduces substantially. Unlike Web of Science, the PatentsView data source provided sufficient metadata to exclude patents with zero references, similarly impacting the data as removing outliers with CD5 = 1 (Fig. 2 and Extended Data Fig. A2). Finally, note again that the curve without CD5 = 1 patents corresponds to (d), the histogram presented in Park et al. [1]"
A screenshot of my Graze feed logic. The sorting is set to "new" and the logic includes a single "all of these" block that contains a "social graph" block set to include only my follows and an "attribute comparison" block where "reply" is set to "null".
Desmond at the podium in front of a slide presentation with the name of the event. The room is formal and has big chandeliers and flags.
A photo of Lauren speaking in front of her slides, titled “Distant Viewing Theory Overview”
Section 3 from the paper. The skills and tools of today’s engineering are very different from those animating photography, film, and the phonograph [1]. Particularly – but not only – in the area of generative AI image production, the disruptions surrounding origin-stories have transformed: while it is true that technology is once again challenging authenticity, the jeopardy this time runs deeper. For Benjamin, technology separated art from its creator. In genAI, there is no historical origin to separate from: nothing has been lost or threatened by the machines because there was nothing there at the beginning. The reason there is no origin is that there is too much origin. And, this excess is a characteristic of how generative AI works technically. When Midjourney or Stability is prompted to create an image, the AI recompiles stacks of paintings stored in its memory, and filters their pixels for features responsive to the prompt. It may be that in that process the Mona Lisa and images of the Mona Lisa, and images of images of Mona Lisa are active somewhere. But where? And to what degree? There are so many squares of color referencing each other that the idea that one or any group could stand out as significant makes no sense. Finding the authentic beginning of a generative AI image would be like trying to find the one grain of sand that first rested on a long beach. It is not an impossible task; it is an irrational one. Of course, it is true that the data training AI generative models does come from the past, and directly or indirectly from specific material objects created by human hands. But even if it is conceded that the Mona Lisa did, in fact, play a significant role in the production of a specific g enAI i mage, then why was i t chosen? Why that painting instead of Rembrandt’s Nightwatch, or one from Caravaggio’s Mathew series? There are no answers to these questions. The magnitude is too high, the number of images called upon to respond to any prompt, the numb…