Lightnews — Scholar-powered news

Reposted by Mercedes 🤦‍♀️ Bunz

Ted Underwood @tedunderwood.com · 13h

You can get totally different advice from a language model depending on how you frame your question. Which could be a weakness. On the other hand, if you're patient enough to ask the question several times, with different framing, you start to get a kind of 3D scan of the possibility space.

4 4 69

Mercedes 🤦‍♀️ Bunz @mrsbunz.bsky.social · 11h

Everyone should try this 😜🧮

‏ deepfates @deepfates.com.deepfates.com.deepfates.com.deepfates.com.deepfates.com · 1d

btw what I did here was I took a data set of ChatGPT interactions collected in the wild and reversed the "assistant" and "user" tags. fine-tuned llama 8B on some of that data and gave it the ability to message you first. 😊

2

Reposted by Mercedes 🤦‍♀️ Bunz

Internet Archive @archive.org · 1d

Libraries, archives & museums are under pressure, but cultural memory can survive.

Hear author @TJOwens.bsky.social of AFTER DISRUPTION: A FUTURE FOR CULTURAL MEMORY chat with @ShannonNattern.bsky.social on the Future Knowledge #podcast.

🎧 Listen & subscribe ⤵️
futureknowledge.transistor.fm

Podcast cover for Episode #11 of 'Future Knowledge' titled 'After Disruption.' The design features portraits of speakers Trevor Owens and Shannon Mattern. Surrounding them are collage-style retro-futuristic elements, including a classical statue head, a vintage computer with the word "NEW" on its screen, a modernist observation tower, and abstract architectural shapes. The title uses bold, blocky typography, giving the cover a vintage sci-fi aesthetic.

1 21 110

Reposted by Mercedes 🤦‍♀️ Bunz

Maria Antoniak @mariaa.bsky.social · 16h

Keynote at #COLM2025: Nicholas Carlini from Anthropic

"Are language models worth it?"

Explains that the prior decade of his work on adversarial images, while it taught us a lot, isn't very applied; it's unlikely anyone is actually altering images of cats in scary ways.

1 21 73

Reposted by Mercedes 🤦‍♀️ Bunz

𝙃𝙤𝙪𝙨𝙚 𝙤𝙛 𝙇𝙚𝙖𝙫𝙚𝙨 Audiobook Narrator @jefferyharrell.bsky.social · 1d

I posted this last night cause I kind of wanted to bury it. I got cold feet about putting it out there.

embedding-space.github.io/sparse-netwo...

The subject is WHY neural networks work, and I think the answer I offer is kind of interesting. Maybe even a little correct, possibly.

A line chart titled “Accuracy vs. Sparsity (Iterative Magnitude Pruning)” showing model accuracy as weights are pruned. The x-axis represents sparsity from 0% to 100%, and the y-axis represents accuracy from 0% to 100%. A blue line with circular markers shows that accuracy stays around 80% from 0% to roughly 90% sparsity, then drops sharply toward 55% near 100% sparsity. A dashed red horizontal line labeled “80% target” runs across the chart near 80% accuracy, indicating the desired baseline.

11 13 82

Mercedes 🤦‍♀️ Bunz @mrsbunz.bsky.social · 2d

This might interest some of you 👉 Doing AI Differently ⚡️sandpit and funding call – open to UK, Canada, US researchers for 4 projects integrating humanities into core AI development, benchmarks + tools that enable AI to handle cultural complexity. Deadline 31 Oct: www.ukri.org/opportunity/...

Expressions of interest: artificial intelligence humanities sandpits: Canada, UK and US

AHRC and SSHRC invite expressions of interest to attend a humanities-led, interdisciplinary research sandpit looking to put humanities insights and methodologies at the heart of artificial intelligenc...

www.ukri.org

1

Reposted by Mercedes 🤦‍♀️ Bunz

The Ass is the father of the legs @internethippo.bsky.social · 6d

"We're going to create superintelligence" How about making outlook search work first. How about that

100 1.7K 10K

Mercedes 🤦‍♀️ Bunz @mrsbunz.bsky.social · 9d

Join us if you are in London to discuss the language of LLMs! Wednesday, 15 October 4pm. 👇

King's Arts & Humanities @kingsartshums.bsky.social · 10d

⌨️ On 15 Oct, Dr @mrsbunz.bsky.social from @kingsdh.bsky.social explores Large Language Models and literary & philosophical theories of language to show that their calculation of meaning is fundamentally unlike meaning produced in communication.

🔗 Read more ⤵️

On The Calculation of Meaning, Or Doing Words Without Things | King's College London

Bringing computational mechanisms of Large Language Models in conversation with literary and philosophical theories of language to show that their calculation of meaning is fundamentally unlike…

buff.ly

1 2

Mercedes 🤦‍♀️ Bunz @mrsbunz.bsky.social · 22d

Oh yes. 🫶👇

Anne Pasek @apasek.bsky.social · 22d

This is good satire: alignmentalignment.ai

Center for the Alignment of AI Alignment Centers

We align the aligners

alignmentalignment.ai

1

Mercedes 🤦‍♀️ Bunz @mrsbunz.bsky.social · Sep 5

Same.

Reposted by Mercedes 🤦‍♀️ Bunz

M. Beatrice Fazi @mbfazi.bsky.social · Sep 3

“First is sight, that is to say the eye; second is the form of the thing seen; third is the distance from the eye to the thing seen …” Piero della Francesca

Spending the day at Galleria Nazionale delle Marche in Urbino 🖼️🎨🏰

2 6

Reposted by Mercedes 🤦‍♀️ Bunz

Ted Underwood @tedunderwood.com · Aug 29

New preprint on "Computational Hermeneutics," co-authored by too many people to list in one post. TL;DR: GenAI is a cultural technology, and needs to be evaluated in ways that recognize situatedness, plurality, and ambiguity as the conditions of meaning — not noise to be minimized.

Computational Hermeneutics: Evaluating Generative AI as a Cultural Technology

<div> <div> <div> <p>Generative AI (GenAI) systems are increasingly recognized as cultural technologies, yet current evaluation frameworks often treat cul

papers.ssrn.com

13 48 140

Reposted by Mercedes 🤦‍♀️ Bunz

Jack Stilgoe @jackstilgoe.bsky.social · Aug 13

We have a new call for papers. If you're working on AI, democracy and public voice, please take a look. You'll need to move fast! digitalgood.net/call-for-abs... @mrsbunz.bsky.social @divya.bsky.social @meredithmeredith.bsky.social

Call for abstracts: Special issue - Including public voices in AI - ESRC Digital Good Network

AI promises to profoundly and pervasively transform the social and natural world. Yet public voices are rarely included in research and development that shapes AI […]

digitalgood.net

1 2 2

Mercedes 🤦‍♀️ Bunz @mrsbunz.bsky.social · Jul 6

Oh no, what a development. Fingers crossed all will be well, I wish you all the time you need for a full recovery! 🦋

Mercedes 🤦‍♀️ Bunz @mrsbunz.bsky.social · Jun 23

I really don’t like our times.

1

Mercedes 🤦‍♀️ Bunz @mrsbunz.bsky.social · Jun 14

… quite enjoyed this conference ‘Chat Token Vector’ 🤖🧮↗️ … with @mttpsq.bsky.social, @hannesbajohr.de, @atg-dbp.bsky.social and many others.

Fabian Offert, Mercedes Bunz, Hannes Bajohr in a Venice patio.

Hand holding up the program of the conference ‘Chat Token Vector’.

1 2 10

Reposted by Mercedes 🤦‍♀️ Bunz

Alexander Doria @dorialexander.bsky.social · Jun 8

Ok I guess I have to go through that Apple paper.

My immediate issue is the framing which is super binary: "Are these models capable of generalizable reasoning, or are they leveraging different forms of pattern matching?" ml-site.cdn-apple.com/papers/the-i...

1 4 44

Mercedes 🤦‍♀️ Bunz @mrsbunz.bsky.social · May 29

Join me at the HBK Braunschweig, Germany next Wednesday June 4 to discuss caring about data and technical collectives. ✊

Poster of an event 4th June in Braunschweig at the Hochschule

2 8

Reposted by Mercedes 🤦‍♀️ Bunz

Ranjodh Singh Dhaliwal @ranjodhdhaliwal.com · May 12

Applications just opened for a postdoctoral position in Digital Media Studies with Markus Krajewski and myself at the Department of Media Studies and the Digital Humanities Laboratory at the University of Basel. (1/3)

1 17 19

Reposted by Mercedes 🤦‍♀️ Bunz

Centre for Technomoral Futures @technomoralfutures.bsky.social · Apr 1

Just over a week until our Technomoral Conversation on AI & Creative Labour!

We’ll be looking at issues ranging from the AI industry’s copyright violations, responses from creatives, and the wider ethical & political questions about the role of AI in creative practice and culture ▶️ edin.ac/3WJhuUH

Blue to green gradient graphic with headshots of the speakers for the Technomoral Conversations event, and text reading: Technomoral Conversations, AI and Creative Labour. Join us on 10 April at 6pm in Edinburgh & online, where we will hear from: Shannon Vallor (University of Edinburgh), Caroline Sinders (University of Arts London, CoRD Labs), Paula Westenberger (Brunel University London), and Richard Combes (The Authors' Licensing & Collecting Society).

1 4 8

Mercedes 🤦‍♀️ Bunz @mrsbunz.bsky.social · Apr 2

How about taking your AI research one year to Rome? The project "Machine Visual Culture: Artificial Intelligence and the History of Seeing" has interesting Predoctoral Fellowships
and Postdoctoral Fellowships 👇

www.biblhertz.it/3628094/2503...

www.biblhertz.it

1

Reposted by Mercedes 🤦‍♀️ Bunz

Hannes Bajohr @hannesbajohr.de · Mar 30

It's out, free and open access:

Thinking _with_ AI: Machine Learning the Humanities.

11 essay on AI and how to think with it – by Peli Grietzer (@peligrietzer.bsky.social), Leif Weatherby (@leifw.bsky.social), Mercedes Bunz (@mrsbunz.bsky.social), Fabian Offert, Lev Manovich…

Open Humanities Press– Thinking <em>with</em> AI

A scholar led open access publishing collective

www.openhumanitiespress.org

1 13 37

Reposted by Mercedes 🤦‍♀️ Bunz

Kate Devlin @drkatedevlin.bsky.social · Mar 26

Applications are now open for 20 interdisciplinary jobs in AI at King’s College London: King’s AI+ Academic Fellowships, advancing AI across *all* disciplines (yes, inc. Arts & Hums — come and work with me!). www.kcl.ac.uk/jobs/role/ki...
@kingsdh.bsky.social @kings-dfi.bsky.social

King's AI+ Academic Fellowships (leading to an open-ended academic post)

King’s is making a major strategic investment into driving AI activity across our nine academic Faculties. Our ambition is to accelerate growth in Artificial Intelligence (AI) and data-driven research...

www.kcl.ac.uk

2 38 49

Reposted by Mercedes 🤦‍♀️ Bunz

Nanna B Thylstrup @nannathylstrup.bsky.social · Mar 24

1 / Executive power now operates via Delete Commands in digital infrastructures. This isn't data management. It's calculated obliteration of who gets to exist in digital memory and knowledge infrastructures. Thank you @verfassungsblog.de for publishing our (@ktmac.bsky.social and my) essay!

Verfassungsblog @verfassungsblog.de · Mar 24

Who decides which data disappears – and what does that mean for democracy?

NANNA THYLSTRUP (nannathylstrup.bsky.social) and KATIE MCKINNON explore how political changes in digital infrastructures reshape access to knowledge and democracy in the U.S. and beyond.

verfassungsblog.de/the-politics...

Quote from our published article by the authors Nanna Bonde Thylstrup and Katie MacKinnon: “The ability to reconfigure digital infrastructure – whether through term removals, restriction of access, or wholesale elimination of databases – constitutes a significant mechanism of control."

1 16 22

Reposted by Mercedes 🤦‍♀️ Bunz

Karen Gregory @karengregory.bsky.social · Mar 24

This looks good:

Computational Social Science meets Qualitative Research
www.lse.ac.uk/Methodology/...

Call for papers

Call for papers, Computational Social Science meets Qualitative Research. Two-day workshop for 10 November 2025 at LSE.

www.lse.ac.uk

5 9