|
Maria Antoniak@mariaa.bsky.social |
756 followers 484 following 366 posts
allen institute for ai • incoming asst prof @ cu boulder (fall 2025) • nlp + cultural analytics + healthcare • political views/opinions my own • maria-antoniak.github.io/
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
A translation I’ve used for a long time is that whenever I see an “ethics” or “privacy” section of a paper that only discusses differential privacy, I interpret this as either “we don’t care much about privacy” or “our corporate bosses/lawyers won’t let us write about real privacy problems.”
0 replies 0 reposts 7 likes
|
Margaret Mitchell
@mmitchell.bsky.social
[ View ] |
Another ethics-informed bit of norm-setting in AI we've started @huggingface.bsky.social :
Providing metadata about the private information in datasets.
huggingface.co/blog/presidi...
Applicable to: Dataset Cards, Auditing, impact assessments, data measurement
0 replies 2 reposts 10 likes
|
Laura K. Nelson
@lauraknelson.bsky.social
[ View ] |
I use almost exclusively 20th and 21st century material in my research and the loops I've gone through to get text mining access over the past 10 years (pre-AI!), or any access at all. Whew it's exhausting, and limits things like posting replication repositories. This is an important perspective 👇
0 replies 2 reposts 10 likes
|
Bellingcat
@bellingcat.com
[ View ] |
Latest: An analysis of open source evidence, as well as multiple missile experts, have pointed to a Russian launched Kh-101 cruise missile being the weapon that struck a children’s hospital in Kyiv www.bellingcat.com/news/2024/07...
9 replies 240 reposts 485 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
Thank you, very interesting!
1 replies 0 reposts 2 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
Do you have any go-to summaries/references for this? It's a topic that comes up a lot in conversation but isn't something I've studied myself.
1 replies 0 reposts 4 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
Two sentences in the entire piece about privacy. We need new healthcare tools, but we also need way more skepticism and scrutiny about tech companies like OpenAI scooping up our personal health information.
1 replies 0 reposts 3 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
she's not like the other girls 😔
0 replies 0 reposts 2 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
Just searched for "python levenshtein distance", how nostalgic, haven't done that in a while.
0 replies 0 reposts 13 likes
|
Scott McGrath
@smcgrath.bsky.social
[ View ] |
🧪 A new study found patients understood radiology reports better with summaries from advanced language models than standard care. Yet, most summaries needed edits, showing human oversight is crucial. 🩺🖥️ 🤖
1 replies 2 reposts 4 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
Turning off replies because is not the place to have this argument.
0 replies 0 reposts 1 likes
|
Computational Humanities Research
@comphumresearch.bsky.social
[ View ] |
📢 #CHR2024 Deadline Alert! 📅 July 8 is just around the corner for paper submissions to the Computational Humanities Research conference in Aarhus. 🏃♀️💨 Polish those drafts or sprint to the finish line! More details at: 2024.computational-humanities-research.org/cfp/ #ComputationalHumanities
0 replies 6 reposts 4 likes
|
Melanie Walsh
@mellymeldubs.bsky.social
[ View ] |
Poetry is weirdly prominent in LLM conversations. But what do models really "know" about poetry?
We tested how well LLMs can recognize 20+ poetic forms in English & probed major pretraining datasets to see which poems might be memorized.
New preprint: arxiv.org/abs/2406.18906
1 replies 17 reposts 39 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
I’m aware, hence the question!
1 replies 0 reposts 1 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
Well, in this case they are.
0 replies 0 reposts 2 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
"LLMs are just next-word predictors." What are the current best references to respond to this, either supporting or critiquing?
6 replies 2 reposts 6 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
Ok but my current greatest organizational innovation is creating a "Today" section in my main Slack workspace, giving it an emoji that indicates how scrambled I am, and then each day moving channels in and out of this section.
1 replies 0 reposts 5 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
Research is complicated, sometimes you need space to do something unexpected, and you’re balancing this against 10 other priorities like TAing. Freedom to manage one’s time is a wonderful luxury and is one reason I chose to do a PhD. Clear communication and organization are still important though!
1 replies 0 reposts 4 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
One more thought that doesn’t seem to be shared: if my advisor had hired a PM and the PM had started tagging me in kanban tickets, I would have done everything to claw my way out of being managed and surveilled. It can be a good system, but it’s one I intentionally left behind in industry.
1 replies 0 reposts 4 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
Yeah this is key. My Slack channels always have lots of bookmarks, and my Google docs usually have a “Quick Links” section where I link to everything (Overleaf, other Google docs, data, etc.). Notion is way better at all of this, of course, but everyone uses Google and no one uses Notion, so.
1 replies 0 reposts 6 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
This might be a feature of bigger collaborations, as keeping a large group up to date requires more managing than when it’s just student + advisor.
1 replies 0 reposts 3 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
On the practical side, when I’m that person, I’ve learned to make and maintain a Slack channel, a meeting notes doc, and a project-specific to-do list.
1 replies 0 reposts 4 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
In my opinion, that’s the role of the “first author” in NLP, the person who is leading the charge and making sure everything that needs to happen happens. That person is usually the PhD student.
1 replies 0 reposts 3 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
Working on this with @mellymeldubs.bsky.social and Anna Preus has been so much fun. Poetry, data, models, evaluation, memorization, and more! 📜🤖
1 replies 1 reposts 5 likes
|
Leslie Root
@lesja.bsky.social
[ View ] |
Love to spend a decade learning how to meticulously quantify human suffering, then get a job doing it, only to find out that, to quote Vonnegut, "the power of this weapon turns out to be that of a custard pie dropped from a stepladder six feet high."
0 replies 2 reposts 12 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
Two sci-fi novels that destroy this perspective, that I read recently and loved:
“I Who Have Never Known Men” by Jacqueline Harpman
www.transitbooks.org/books/iwhoha...
“The Wall” by Marlen Haushofer
bookshop.org/p/books/the-...
1 replies 1 reposts 4 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
Interdisciplinary researchers, please take our survey!
▶️ forms.gle/XULo3xTbkm9n...
Just 6 questions, should take just 2-3 minutes, and we would be so grateful for your input 🙏
Calling digital humanists, computational social scientists, and anyone doing research in multiple communities!
0 replies 14 reposts 10 likes
|
|
Maria Antoniak
@mariaa.bsky.social
[ View ] |
“The widespread integration of Gemini with Google search and GPT-4.0 with Bing search has led to a shift in the trend of self-diagnosis… [comparing] GPT-4.0 and the Gemini model on the task of self-diagnosis… [we find] accuracies of 63.07% and 6.01%, respectively.”