Sign in

DM Howcroft 🦔

@davehowcroft.com
536 followers 537 following 1.2K posts

Computational linguist teaching computers to use their words in Edinburgh. Queer. Catholic. Immigrant. they/them Languages: en, de, fr, sv Also available at mastodon.social/@_dmh

PostsRepliesMedia
Reposted by DM Howcroft 🦔
Jess Maddox @jessmaddox.bsky.social · 29/09/2026
Apparently when you yap on the internet for long enough, they let you co-host a podcast. Thrilled to announce I’m joining BBC’s The Interface, their podcast on all things tech. www.bbc.co.uk/mediacentre/...
bbc.co.uk
The Interface, BBC Studios' flagship technology podcast, returns with a new co-host and refreshed editorial focus
This season, Jessica Maddox joins returning co-host and BBC technology columnist, Thomas Germain
071
Reposted by DM Howcroft 🦔
Benjamin @benjaminschneider.ch · 9h
this is what your statements turn into if you add too many hedges
1787
Reposted by DM Howcroft 🦔
Nina Markl 🏳️‍⚧️🏳️‍🌈 @ninamarkl.bsky.social · 10h
after citing some work from EMNLP 2024 with page number 13546, i went to check how many papers they've been publishing and, uh, it's gone from 800 in 2020 to 1800 in 2025??
262
Reposted by DM Howcroft 🦔
vortex_egg @vortexegg.com · 28/09/2026
The push to make it so that "websites as a destination for humans is over" is an intentional move to violate the freedom to read of individuals and move access to information into a "pay to get an opaquely intermediated version from a nontransparent source" model.
716657
Reposted by DM Howcroft 🦔
Sketch Engine @sketchengine.eu · 29/09/2026
⏰ Deadline tomorrow. Built a corpus, dictionary, or language tool? Apply for the Adam Kilgarriff Prize by 30 September. #CorpusLinguistics #NLP kilgarriff.co.uk/prize/
001
Reposted by DM Howcroft 🦔
Tressie McMillan Cottom @tressiemcphd.bsky.social · 29/09/2026
Believe it or not, a lot of people do not have to-do lists. And the the lists many of them have are not structured for an ai agent to be helpful. This isn’t groundbreaking. If you have ever watched the evening news or read even a single story about people who aren’t like you, it is pretty basic
3085097
Reposted by DM Howcroft 🦔
Erin Biba @erinbiba.bsky.social · 29/09/2026
I’ve had so many Christians absolutely shocked that as Jews we don’t worship Jesus. I’ve been met with utter confusion. “Ok so how does he factor then? What do you see him as?” “We don’t even think about him. He was just a random guy.” They refuse to believe me.
3250350
Reposted by DM Howcroft 🦔
Dallas Card @dallascard.bsky.social · 28/09/2026
Amazing postdoc opportunity!
001
Reposted by DM Howcroft 🦔
Jess Maddox @jessmaddox.bsky.social · 28/09/2026
As an editor-in-chief of a new academic journal, your hallucinated, AI slop submissions risk damaging our credibility. Your AI use has an impact beyond you.
081
Reposted by DM Howcroft 🦔
Tressie McMillan Cottom @tressiemcphd.bsky.social · 28/09/2026
In practice, AI literacy doesn’t start with building the tool for your personal use. It starts with understanding yourself as a collection of tasks who should subcontract those out to a machine. Almost no regular person’s life currently exists as such.
9825209
Reposted by DM Howcroft 🦔
Tressie McMillan Cottom @tressiemcphd.bsky.social · 28/09/2026
I am talking about working with people to even conceive of their lives as a set of tasks — that isn’t exactly natural and isn’t something we train people to do and yet its a precondition for figuring out what an agent would even *do*.
1199050
Reposted by DM Howcroft 🦔
Ehud Reiter @ehudreiter.bsky.social · 28/09/2026
I see lots of posts "our lab has N papers in VENUE". Congratulations, but what is more useful to me is short 3-5 line descriptions of a paper which will convince me to open it and read abstract. With link to final (not submitted) version
101
Reposted by DM Howcroft 🦔
Spencer Ackerman @attackerman.bsky.social · 26/09/2026
Ha ha that’s gross and should make everyone uncomfortable, anyway, Rahm presided over a Chicago police black site, lied about what it was and then I sued the Chicago police for its records that proved it:
theguardian.com
Homan Square revealed: how Chicago police 'disappeared' 7,000 people
Exclusive: Lawsuit exposes scale of detention at off-the-books interrogation warehouse while attorneys describe find-your-client chase ‘from a Bond movie’
9147371723
Reposted by DM Howcroft 🦔
Laura K. Nelson @lauraknelson.bsky.social · 25/09/2026
You are talking to a bright, eager undergraduate student in CS. They are convinced that scientific reasoning is not socially/culturally conditioned. That is, the scientific method is the scientific method, whether it's 2026 Canada or 1850 China. What would you recommend they read? (Note the CS & UG)
15162
Reposted by DM Howcroft 🦔
justine zhang @tisjune.bsky.social · 24/09/2026
they're saying the AI can do math and science again, so it's time to reread Thurston's On Proof and Progress in Mathematics, a lovely articulation of the social work of assembling understanding, i.e., the best reason to tackle hard problems. www.ams.org/journals/bul...
When I started as a graduate student at Berkeley, I had trouble imagining
how I could "prove" a new and interesting mathematical theorem. I didn't
really understand what a "proof was.
By going to seminars, reading papers, and talking to other graduate students,
I gradually began to catch on. Within any field, there are certain theorems and
certain techniques that are generally known and generally accepted. When you
write a paper, you refer to these without proof. You look at other papers in the
field, and you see what facts they quote without proof, and what they cite in
their bibliography. You learn from other people some idea of the proofs. Then
you're free to quote the same theorem and cite the same citations. You don't
necessarily have to read the full papers or books that are in your bibliography.
Many of the things that are generally known are things for which there may be
no known written source. As long as people in the field are comfortable that
the idea works, it doesn't need to have a formal written source.At first I was highly suspicious of this process. I would doubt whether a certain idea was really established. But I found that I could ask people, and they
could produce explanations and proofs, or else refer me to other people or to
written sources that would give explanations and proofs. There were published
theorems that were generally known to be false, or where the proofs were generally known to be incomplete. Mathematical knowledge and understanding were
embedded in the minds and in the social fabric of the community of people
thinking about a particular topic. This knowledge was supported by written
documents, but the written documents were not really primary.
411237
Reposted by DM Howcroft 🦔
Jordan Boyd-Graber @boydgraber.bsky.social · 25/09/2026
Lab meetings recently.
Muppets in back seat of car: Prof, can we get Jev?
Kermit, driving: We have Jev at home, on the server.
Classifier at Home: A picture of BERT staring intently at paper clips.
1202
Reposted by DM Howcroft 🦔
beka valentine @bekavalentine.bsky.social · 25/09/2026
anyone who knows the long history of Silicon Valley knows that collaboration with the military is deeply rooted in SV, from the post-WW2 radar boom era to the Fairchild transistor era and beyond. frankly, Silicon Valley, and indeed computers/the net as a whole, exists how they do because of it
1106
Reposted by DM Howcroft 🦔
Łukasz Langa @lukasz.langa.pl · 05/08/2026
You can help shape the future of #Python typing with the power of having an opinion! Fill out the annual Python Type System and Tooling Survey: www.surveymonkey.com/r/python_typ...
surveymonkey.com
Python Type System and Tooling Survey 2026
The Python Typing Survey offers a comprehensive look at the current state of Python’s type system and developer tooling. It captures the evolving sentiment, challenges, and opportunities around Python...
275
DM Howcroft 🦔 @davehowcroft.com · 25/09/2026
Reminder of this excellent free reference resource!
000
Reposted by DM Howcroft 🦔
Ehud Reiter @ehudreiter.bsky.social · 25/09/2026
Excited that my student Yujun Wang will present (oral) at EMNLP on "Beyond Accuracy: Community Perspectives on Machine Translation". Many of my students have looked at stakeholder requirements/concerns. Need to understand in order to align AI with humans! arxiv.org/abs/2606.09655
arxiv.org
Beyond Accuracy: Community Perspectives on Machine Translation
Despite remarkable progress in machine translation (MT), non-AI communities have raised growing concerns about MT systems, suggesting a noticeable gap between technical advancement and the needs of re...
091
Reposted by DM Howcroft 🦔
Laurel Brehm @drlearnasaurus.bsky.social · 24/09/2026
Columbia psych is hiring
043
Reposted by DM Howcroft 🦔
Homestar Blade Runner 20X6 @checarina.bsky.social · 24/09/2026
fuckin THANK you
0214
Reposted by DM Howcroft 🦔
Alex Hanna @alexhanna.bsky.social · 23/09/2026
“AI” IS impacting labor, but not in the ways you think. In some cases, people generating “workslop” are making their colleagues’ jobs more difficult. Snippet from my video Myths of AI and Productivity, in which we at dispel 5 common myths around AI and labor. www.youtube.com/watch?v=3BO0...
There are three stacked screenshots from a video of Alex, a brown trans woman, sitting on a pink chair and speaking into the camera while holding a small microphone. She has short black curly hair and is wearing a blue long-sleeved shirt and light tan pants. Behind her is a cat-shaped pillow with an image of a tuxedo cat on it, and behind her chair are rows of books on a light blue wall. Underneath are captions from the video that say: “But generative AI can also reduce wages in other ways, namely by punishing workers who use the tools and pass on ‘workslop’ to their colleagues.”, “...researchers report that receiving generative AI text or images from colleagues results in lower perceptions of the competence of their colleagues.” and “Getting slop from your colleague makes you think worse of them. And if those workers are women or older, or people of color, then it confirms their co-workers’ biases against them.”
814041
Reposted by DM Howcroft 🦔
Shannon Vallor @shannonvallor.bsky.social · 24/09/2026
In any reasonable society the next sentence would be ‘and so they were all put in jail for admitting to gross negligence and malfeasance on a scale previously unthinkable’
212243
Reposted by DM Howcroft 🦔
Melissa Terras @melissaterras.bsky.social · 23/09/2026
Finishing up the @transkribus.bsky.social user conference. I’ve been working on this for over 13 years now and I am so, so proud of the infrastructure, community, and team delivering our ATR and information extraction tools, cooperatively. A strong sense of community here! 🦾
We Have Built Unique Al in a Sea of AI
Hundreds of Features
Thousands
of Projects
Tens of
thousands of Models
Hundreds of thousands
of Users

Slide presentation
3172
DM Howcroft 🦔 @davehowcroft.com · 23/09/2026
Apparently @brightonuni.bsky.social is planning to derecognize their employees' unions!? Way to say you don't care about your staff's working environments or student's learning environments! Make sure you express your support for @brightonucu.bsky.social www.ucu.org.uk/article/1464...
ucu.org.uk
UCU condemns University of Brighton derecognition plans
UCU has today condemned plans by the University of Brighton to derecognise UCU on campus.
011
Reposted by DM Howcroft 🦔
Curtis @soccurt.bsky.social · 23/09/2026
We're just going to supercharge every Onion article with AI aren't we theonion.com/black-guy-ph...
theonion.com
Black Guy Photoshopped In
AMES, IA–In the spirit of celebrating diversity at Iowa State University, a black guy was digitally added to the cover of the school’s 2001 spring-semester course catalog, school officials announced M...
031
Reposted by DM Howcroft 🦔
David Mimno @dmimno.bsky.social · 22/09/2026
It's possible for Jev/Laya/Decision Models to be not that big a deal as tech and massive as a new paradigm. Here's why I'm really excited from an NLP history perspective (thread)
17619
Reposted by DM Howcroft 🦔
DM Howcroft 🦔 @davehowcroft.com · 23/09/2026
I think one of the issues we're seeing is a disconnect between busyness/output and productivity--I think productivity assumes there is value in the product and it will be a while until the majority of "paper shaped objects" are in fact papers.
033
Reposted by DM Howcroft 🦔
Ehud Reiter @ehudreiter.bsky.social · 23/09/2026
AI is improving productivity in science, but so far this has led to a flood of low-quality spam and large amounts of so-so papers. I dont see increase in the really high-quality research that science needs! Maybe blame perverse incentives in science that priotitise quantity.
142
Reposted by DM Howcroft 🦔
Clément Canonne @ccanonne.github.io · 22/09/2026
"If OpenAI and Anthropic really want to help advance science, there is one obvious way they can do so: They can spend a tiny fraction of their vast fortunes to fund the kind of humble but essential research that’s increasingly starved for resources." (That they don't, and won't, says it all)
1285
Reposted by DM Howcroft 🦔
I Travel Like Sound Dog @chadstanton.blacksky.app · 22/09/2026
The outright refusal to treat white supremacy as an ideology which motivates action and treat it as generally objectionable is what allows this strategy to be viable for Elon Musk.
2532441055
Reposted by DM Howcroft 🦔
Ehud Reiter @ehudreiter.bsky.social · 22/09/2026
If AI is risk to humanity, because of human stupidity, not clever AI. In 1983 hallucinating Soviet satellite almost started nuclear war, stopped by sensible human. In 2026, hallucinating AI almost led US military to attack Chinese ship, due to poor human oversight edition.cnn.com/2026/09/18/p...
edition.cnn.com
Exclusive: US military had close call after using AI for false intelligence report, sources say | CNN Politics
The episode shows the risks of using this new, relatively poorly understood technology in the middle of the Iran war
021
Reposted by DM Howcroft 🦔
Mar Hicks @histoftech.bsky.social · 19/09/2026
Precisely this—the power attributed to technology makes it incredibly useful for laundering responsibility
12539183
Reposted by DM Howcroft 🦔
Shiv Ramdas Mens Rice Activist @nameshiv.bsky.social · 21/09/2026
on one hand this is good pushback and is needed. On the other hand, the political press being willing to play good lapdog and nod along while he abused literally everyone till such time as he started beating them too tells us a lot about the character of the people in that pool
271517247
Reposted by DM Howcroft 🦔
Dr Abeba Birhane @abeba.blacksky.app · 21/09/2026
"In a leaked document, the Irish Presidency proposes to EU Member States that the interest of OpenAI, Anthropic, Google, Meta and SpaceX to make profits should take precedence over the fundamental right to data protection" noyb.eu/en/ai-eu-mem...
noyb.eu
AI: EU Member States plan “digital expropriation” of Europeans in the interest of AI companies
In a leaked document, the Irish Presidency proposes to EU Member States that the interest of AI companies to make profits should take precedence over the fundamental rights of Europeans.
18464294
Reposted by DM Howcroft 🦔
Marcel Bollmann @marcel.bollmann.me · 21/09/2026
Did you know it’s pretty easy to programmatically access ACL Anthology data? For example, here’s a snippet you could use to get the BibTeX entries for all of your papers. The downside is that the initial fetching of the data takes quite long atm, but it can still be useful for automation!
Python code:

from acl_anthology import Anthology
anthology = Anthology.from_repo()
author = anthology.get_person("marcel-bollmann")
for paper in sorted(author.papers(), key=lambda p: p.year):
    print(paper.to_bibtex())
3164
Reposted by DM Howcroft 🦔
Dave Fernig @davefernig.eurosky.social · 20/09/2026
Absolutely true, as it is for Roman cement, recipe lost for >1500 years which is miraculously self-healing despite its inability to understand anything.
093
Reposted by DM Howcroft 🦔
I Travel Like Sound Dog @chadstanton.blacksky.app · 20/09/2026
It was a perfect retort, instead of acting as if these people have any serious position to consider just a quick concise “get bent loser”
1101
Reposted by DM Howcroft 🦔
Dr. Garrett Schumann @garrt.bsky.social · 20/09/2026
I know I'm not the first to observe this, but "get a life" is such a good counter to current BS culture skirmishes, it was such a common phrase growing up in the 90s and I assume its retreat from common discourse enabled, in some way, the arrival of the present, problematic media environment
2373
Reposted by DM Howcroft 🦔
Tressie McMillan Cottom @tressiemcphd.bsky.social · 20/09/2026
👇🏾👇🏾👇🏾👇🏾👇🏾👇🏾
017132
Reposted by DM Howcroft 🦔
Ehud Reiter @ehudreiter.bsky.social · 20/09/2026
love the phrase "paper-shaped objects" "modern AI systems can generate paper-shaped objects even for users who are unfamiliar with the related work and writing conventions for a given area, and who are unable to understand or verify any of a paper’s main claims." blog.iclr.cc/2026/09/02/s...
blog.iclr.cc
Submission policies for ICLR 2027 – ICLR Blog
052
Reposted by DM Howcroft 🦔
Rebecca Sear @rebeccasear.bsky.social · 19/09/2026
If you read the replies you’ll see that Pinker was always a race science supporter, it’s just that nobody paid any attention. Maybe we should start noticing when celebrity scientists are fans of pseudoscientific ideas, and adjust our opinion of their competence accordingly
22854191
Reposted by DM Howcroft 🦔
km fedyk 🐀 @karigrafia.bsky.social · 18/09/2026
Antek, obligatory reading in primary school, a novella about hardships of 19th c rural life. Includes a scene where a girl is burned alive in the oven to cure her fever. I wish I could forget it.
231
Reposted by DM Howcroft 🦔
Jan Vanhove @janhove.bsky.social · 18/09/2026
New blog post: "Cluster analysis: A skeptic’s guide". In which I ask a few questions that social scientists wishing to identify hidden classes in their data ought to address. janhove.github.io/posts/2026-0...
A two-dimensional scatterplot showing a three-cluster solution.

Caption: "Figure 2: A visualisation of a Latent Profile Analysis fit. Such visualisations may help readers appreciate that the clusters aren’t nicely separated and that the researchers’ notion of clusters may not correspond to their own."Table of contents:

Refresher: What is cluster analysis?
What are the clusters for?
What clusters, exactly?
Does the pipeline work?
What does the solution look like?
When running follow-up analyses, how is the uncertainty in the cluster assignments taken into account?
Was it worth it?
Conclusion
References
416145
Reposted by DM Howcroft 🦔
Language on the Move @languageonthemove.bsky.social · 19/09/2026
“AI allows you to get to a bad idea faster”
012
DM Howcroft 🦔 @davehowcroft.com · 19/09/2026
Lol, massive job cuts looming at @edinburgh-uni.bsky.social, but funds for two temporary hired guns to come spend 2 years "managing change". Wonder how much their consultants charged them to advise adding these roles...
Job listings for two change manager roles at university of Edinburgh, one of them senior. Salaries from 41 to 61 thousand.
001
Reposted by DM Howcroft 🦔
Yael Schacher @yaelschacher.bsky.social · 18/09/2026
You can’t just deport people to third countries because you paid off that country’s government for an assurance no rights violations will occur. That’s just not what US law says and not how the Refugee and Torture conventions work. storage.courtlistener.com/recap/gov.us...
storage.courtlistener.com
Main Document – #00108510897 in D.V.D. v. U.S. Department of Homeland Security (1st Cir., 26-1212) – CourtListener.com
OPINION issued by Lara E. Montecalvo, Appellate Judge; Jeffrey R. Howard, Appellate Judge; and Seth Robert Aframe, Appellate Judge. Published. [26-1212] (AVN) [Entered: 09/18/2026 03:22 PM]
24821