Tessera
Folgen

Mercedes 🤦‍♀️ Bunz

@mrsbunz.bsky.social

King’s College London. Interested in reading ML as a ‘calculation of meaning’ 🧮📝📚✍️🎨. Creative AI Lab with Serpentine’s Eva Jäger. Left AI ✊

110 Posts1173 Follower327 Folgt
Eryk Salvaggio@eryk.bsky.social · 4w ago

If the AI / Human intelligence distinction seems blurry it’s because the language is identical but emerges from ultimately incomparable systems. Our presence resolves the gap and makes it seem otherwise — language is always unifying the system instead of creating space to recognize the division.

Pekka Lund@pekka.bsky.social · 4w ago

This is really interesting. Both because of what it reveals about internal processing in LLMs (other than Claude too), and response by GNW pioneers below. "We view this finding as a landmark in consciousness research, because it provides a mechanistic, testable version of the GNW hypothesis."

Stanislas Dehaene@standehaene.bsky.social · last mo.

The global neuronal workspace (GNW) is currently the best documented neuroscience mechanism by which conscious processing arises in the human brain — and now Anthropic researchers have discovered a similar workspace inside their large language model ! www.anthropic.com/research/glo...

Mark Peifer (He, him)@peiferlabunc.bsky.social · last mo.

Truly bizarre story of how two papers from the 1940s by physics pioneer Max Planck have been "retracted" from Naturwissenschaften, potentially by an algorithm, for anachronistic "plagiarism" issues, leaving the current editor baffled (but it is Springer Nature so.... www.science.org/content/arti...

Why have papers by one of history’s most famous physicists been retracted?

Springer Nature has removed two studies by Max Planck. A bot may be to blame

science.org

Mercedes 🤦‍♀️ Bunz@mrsbunz.bsky.social · last mo.

This blogpost by @kripken.com has some good points. 👇 It argues: while there are reasons to call out wrong mysticism of the AI hype, AI models do have computational mechanisms that we still need to be scientifically curious about and would like to understand better kripken.github.io/blog/neurosc...

We Don't Understand Neural Networks At The Algorithmic Level

The largest ongoing debate about AI is “Are Large Language Models (LLMs) intelligent?” That makes sense, at least: the evidence is ambiguous and the stakes a...

kripken.github.io

Sarah Kay@k-interarma.bsky.social · 2mo ago

Aerial PARTIAL view of the anti racist, anti fascist crowd in Belfast right now. City Hall to the left for scale. This - this is the real Belfast. Don’t let anyone tell you otherwise.

Bild
Michael Okun@michael-okun.bsky.social · 2mo ago

I’ve officially resigned as Associate Editor for Frontiers in Systems Neuroscience. It used to be a reputable journal, but became a case study in how forced automation destroys academic integrity. 👇

Ted Underwood@tedunderwood.com · 2mo ago

Psst, buddy — fixer 4 is back, and it shipped the wire-up not the cop-out. Dodge a collision and never ride exports: the local badge scoped truthfully.

Bild
Ethan Mollick@emollick.bsky.social · 2mo ago

One thing I mentioned in passing in my Fable post is that, for long running tasks, Fable starts to develop its own dialect as its many agents and tasks reinforce themselves and make Claudish language ever more Claudish. You need to ask it to report out in plain English or this happens after 9 hours

Mercedes 🤦‍♀️ Bunz@mrsbunz.bsky.social · 3mo ago

Come join us discussing LLMs, language and the humanities on Thursday, June 11 5.30pm UK time. It's my belated inaugural lecture 😱 so I am not coming alone - thankfully @joannazylinska.bsky.social will be with me. 👇 www.tickettailor.com/events/kings...

Register – Culture as Calculation: Technology, Language, and the Machine Learning Turn in Times of Polycrisis – Nash Lecture Theatre (KIN 205)

Culture as Calculation: Technology, Language, and the Machine Learning Turn in Times of Polycrisis – Nash Lecture Theatre (KIN 205), Thu 11 Jun 2026 - Inaugural Lecture: Mercedes Bunz, Professor in Di...

tickettailor.com

Mercedes 🤦‍♀️ Bunz@mrsbunz.bsky.social · 3mo ago

This is good! I might have overlooked others but finally a text running through the history of LLMs exploring the textual mechanics that moved LLM systems from creating language structures to generating relevance👇

John David Pressman@jdp.extropian.net · 3mo ago

I have written a new post, "Implications Of Predicting The Next Token" in which I explain why saying a language model "just" predicts the next token fails to understand what that phrase would even imply. minihf.com/posts/2026-0...

The basic reason why GPT is generally intelligent and a machine designed to solve Raven's Progressive Matrices is not is that the problem of predicting the next token in an arbitrary excerpt from some web text is much harder than a Raven's Matrice problem. Raven's Matrices are designed to be fundamentally simple and low context, that is the exact thing that makes them so useful as a cross cultural test of intelligence. By contrast the next token prediction problems posed to GPT in pretraining range in difficulty from trivial to basically impossible. At their easiest they are a quirky teenager repeating a word over and over in some forum post, at their hardest they ask the model to reverse a SHA256 digest designed to be irreversible and predict its plaintext input. Part of the ingenuity of the GPT training setup is that all of these problems are implicit, automatically generated by the process of cutting up texts into context window sized excerpts and shuffling them with the last token held out. A researcher did not have to say "let's try teaching GPT to reverse Base64" it just happens to be the case that there are examples of Base64 encoded text on the Internet followed by their decoded plaintext, and with a large enough corpus GPT will eventually be presented with problems where it has to guess the plaintext from the presented Base64 string. Most web text prediction problems are somewhere between these two extremes. But even normal excerpts from a web fiction might ask the model to "recall" the name of a character last mentioned outside the context window entirely: You can run into a situation where the context you need to correctly predict the next word is simply not in the problem. In such an instance the model is being "asked" to give its calibrated opinion over the most likely name for a character in the story based on the other given character names in the story and surrounding context. The model is generally intelligent because it is being asked to put down a…
Mercedes 🤦‍♀️ Bunz@mrsbunz.bsky.social · 3mo ago

In London Wed, 3 Jun and no plans for 5pm? My wonderful colleague @joannazylinska.bsky.social ◼️ is presenting her film 📽️ on collapse 🫠 🤖 supported by @slehuede.bsky.social, @eleonoralima.bsky.social + me (not that she needs support). Join us! 👇https://www.tickettailor.com/events/dfistory/2193124

Select tickets – Collapse: A Story – S-2.23 - Strand Building

Collapse: A Story – S-2.23 - Strand Building, Wed 3 Jun 2026 - Collapse – a state of events in which multiple systems fail at once, without the promise of repair – is one of the key terms used today t...

tickettailor.com

Philip N Cohen@philipncohen.com · 3mo ago

Two new facts stand out: 1. 85% of hallucinated citations in preprints are also in the subsequent journal version (thanks, peer review!) 2. Fake cites more likely to use the names of (male) scholars who are already highly cited, creating a fake-citation Matthew effect. arxiv.org/abs/2605.07723

LLM hallucinations in the wild: Large-scale evidence from non-existent citations

Large language models (LLMs) are known to generate plausible but false information across a wide range of contexts, yet the real-world magnitude and consequences of this hallucination problem remain p...

arxiv.org

David Timoney@fromarsetoelbow.bsky.social · 3mo ago

Sociology regularly reveals this paradox: that people will vote for rightwing disruptors not because they share their politics but because they are fed up with the system. It's almost as if the system generates its own opposition in a guise that ensures no actual disruption.

Bild
United Tech & Allied Workers@utaw.tech · 3mo ago

BREAKING: Google DeepMind workers in UK have voted 98% in favour of unionising — demanding an end to AI use by Israel and the US military for perpetrating genocide in Gaza. If successful, they’ll be the first frontier AI lab in the world to unionise.

GOOGLE DEEPMIND WORKERS IN UNION RECOGNITION BID - 98% of Google DeepMind members vote YES
Mercedes 🤦‍♀️ Bunz@mrsbunz.bsky.social · 3mo ago

Pls apply! 👇 My collaborator Serpentine Arts Tech opened 6 a months fellowship for artists, curators, technologists, organisers etc. exploring the breakdown of boundaries that before kept technical systems apart www.serpentinegalleries.org/whats-on/fut...

Future Art Ecosystems R&D Fellowship: Art x Convergence | Serpentine

Serpentine Arts Technologies launches the inaugural Future Art Ecosystems (FAE) R&D Fellowship: Art x Convergence.

serpentinegalleries.org