Willem Zuidema. Associate Professor of Natural Language Processing, Cognitive Modelling & Explainable AI, at the Institute for Logic, Language & Computation, University of Amsterdam.
Want to learn how to analyze the inner workings of speech processing models? š
Check out the programme for our tutorial, taking place at this year's Interspeech conference in Rotterdam: https://interpretingdl.github.io/speech-interpretability-tutorial/
The schedule features presentations and interactive sessions with a great team of co-organizers: Charlotte Pouw, Gaofei Shen, Martijn Bentum, Tom Lentz, @hmohebbi, @wzuidema, @gchrupala (and me!). We look forward to seeing you there š
A.s. donderdag spreekt de Commissie Digitale Zaken met diverse stakeholders over digitale soevereiniteit. Experts @bert_hubert Hubert, Reijer Passchier en Paul Timmers maken zich ernstige zorgen.
https://ibestuur.nl/artikel/actie-nodig-om-de-digitale-soevereiniteit-van-nederland-en-europa-te-versterken/
"As the fediverse continues to grow and evolve, publishers are setting up shop in this new ecosystem. Some outlets have a fediverse strategy that complements their continued activity on traditional social media channels. Others have chosen to abandon traditional channels entirely in order to build a presence thatās more aligned with their values on the open social web."
Elon Musk is for America what Cecil Rhodes was for the British Empire: an oligarch with far-reaching powers bestowed on him to help the state to grab as much of the worldās waterways, land, resources and labor as it can
My latest for @foreignpolicy
https://foreignpolicy.com/2025/02/07/elon-musk-predatory-capitalism-colonialism-cecil-rhodes/
Dit weekend in de Volkskrant: stuk van @laurensvhg over GPT-o1 en de 'briljante marketing' (@wzuidema) van OpenAI https://www.volkskrant.nl/tech/het-nieuwe-ai-model-van-chatgpt-zou-denken-als-een-mens-vooral-een-staaltje-briljante-marketing~b1e494c7/
@kimvsparrentak Ik ben erbij! Benieuwd naar de discussies, en ik baal al bij voorbaat dat ik niet bij @bert_hubert in de parallel-sessie kan zijn...
my current thoughts on the Bluesky boom: good for them!
My time here has brought home to me how much people want very different things from micro blogging: some want a feed of fast moving, short, quippy, posts with reach (essentially early Twitter), others want slower, thoughtful (and maybe kinder) discussion, with privacy and control
if that splits into different platforms with different culture (and design), that may ultimately be helpful, and if the two systems have some degree 1/2
@Rachel_Thorn It sounds like you're looking for an activist rather than an expert... Japan has many excellent Natural Language Processing experts, and even if they are fascinated by LLMs and not simply 'anti', they might be a more worthwhile addition to your symposium than you give them credit for. If I were preparing for the impact of a tornado or hurricane, I'd rather want a proper meteorologist at my symposium (even one fascinated about storms!) than an anti-hurricane activist.
āØ Do current neural speech models show human-like linguistic biases in speech perception?
We took inspiration from classic phonetic categorization experiments to explore whether & where sensitivity to phonotactic context emerges in Wav2Vec2 models š
(w/ @wzuidema )
š
https://arxiv.org/abs/2407.03005
ā¬ļø
Gefeliciteerd - heel benieuwd wat het gaat brengen!
LinkedIn geeft intussen nog wat nuttige extra insights :).
@pbloem Some serious subtooting happening here.
Not just "she says, he says" but an investigation of who has the better arguments. Not just professional AI-debaters, but people that actually do research on AI or on the impact it has. Not just "Hinton predicts the end of humanity", but some serious detail on how negative consequences may come about. Not just "there are optimists, pessimists and skeptics", but an analysis of who profits from all the hype, and who may be deliberately fueling it and why.
2/2
Imagine being "chief features writer" for the Financial Times, and for your feature on Artificial Intelligence hype you paste together a couple of quotes from the usual suspects you know from Twitter: Marcus, Bender, Chollet, ...
Henry Mance gets away with it, but I think the debate on AI & society deserve a wider set of voices, and journalists that dig a little deeper. 1/2
https://www.ft.com/content/648228e7-11eb-4e1a-b0d5-e65a638e6135
@jasmijn02 heftich!
@deevybee "guest authorship" is shockingly common, I'm afraid, to the extent that many established academics don't even see a problem anymore with having their name on papers without any real contribution to the content, and many early career researchers think it's just the way it works.
One quibble with the linked article, though: if asking "occasional incompetent questions" is already an obstacle to authorship, many of us --me included-- can't coauthor anything anymore... š
In a new paper, published today in Current Biology, we analyse the genome of renowned composer Ludwig van Beethoven using a polygenic index related to musicality, as a way to illustrate the limits of genetic predictions at the individual level. Beethoven, one of the most celebrated musicians in history, scored unremarkably, ranking between the 9th & 11th percentile based on modern samples. We explain why this is no surprise & how it can provide a valuable teaching moment on the complex relationships between DNA & behaviour.
An interdisciplinary collaboration across two Max Planck Institutes (Psycholinguistics in Nijmegen & Empirical Aesthetics in Frankfurt), University of Amsterdam, Karolinska Institute, Vanderbilt University and others.
#MastodonScience #science #music #genetics #genomics
@mpi_nl @maxplanckgesellschaft
Haha, here's the commentary in full, a lot shorter than I expected. Content-wise, I rate my predicions as 2/3 correct:
(1) Not explicitly there, suprisingly. They just say infants can learn from less data.
(2) Spot on: point three of the comment.
(3) Implicit in point (1) of the commentary: language is hierarchical, LLMs only probabilistic.
As usual (but length is a good excuse, just this time), the commentary doesn't engage with the neural network literature aware of all those arguments.
(1) LLMs cannot learn subtle syntactic constraints. Illustrated with an example where they fail to complete a complex sentence (e.g., with a double embedding) or fails to say which one of two is grammatical.
(2) LLMs do not reproduce human inabilities to learn certain formal structures. Illustrated with an example from a programming language that LLMs learn perfectly, but naive human subjects fail at.
(3) This should be no surprise, because LLMs are next-word predictors that cannot do *merge*.
I just got a notification about this commentary published in Nature: "Three reasons why AI doesnāt model human language", by Johan J. Bolhuis, Stephen Crain, Sandiway Fong & Andrea Moro.
I have seen so many commentaries by this team and their various coauthors (including, most prominently, Noam Chomsky) that I think I can spell out what they are going to say *before* reading the paper. So here are my three pregistered predictions: 1/3
