For all del, men det handler om hvor mye tid vi skal vie til forskjellige aspekter. At maskiner oppnår super-intelligens er etter min mening et smalt tema oppi det hele.

Jeg syns vel hypotetiske fremtids sci-fi scenarioer er mindre interessante enn reelle utfordringer med bruk av LLMer idag. Se for eksempel www.media.mit.edu/publications...

veldig rar burn av Sterri her i søsterartikkelen om Innsikts engasjement i den offentlige debatten om KI.

(+) Inga Strümke: – Spekulativ skit fortrenger viktige diskusjoner Skrekkscenarioene fra Silicon Valley og tankesmien Langsikt bidrar til PR for KI-bransjen og skygger for viktigere diskusjoner, mener Inga Strümke.

Flott av Inga Strümke i Morgenbladet, setter ord på mye som jeg selv har tenket på det siste året, og som jeg tror mange andre som jobber med ML fra matematikk / statistikk siden tenker.

I think Mistral's Le Chat should have a Proust mode by default

From someone who made the switch sociology->maths, there is a really odd «methodology hierarchy» in parts of social science. For some reason quantitative is deemed «harder than» qualitative, therefore better — total nonsense of course. Methods should adapt to research questions. Stick to your guns!

New statistical methods shed light on medieval literary mystery - Department of Mathematics Full of adventures, battles and love stories, the chivalry romance Tirant lo Blanch is a masterpiece of medieval literature. The novel, sometimes claimed to be the world's first, is made even more fas...

Very interesting, and congrats to the fresh doctor! this reminded me of this blog post from a while back viewing this as a changepoint detection problem. Trying to figure out when in the text author 2 took over from author 1

Nice.

Show us the figure!

Yeah, I'm sure there is literature here, I just need to dig a bit! I've been doing some quite crude thresholding so far to decide when things are low-rank 'enough' for the switch. This works sometimes, but seem to also interfere with adaptation of my HMC sampler

Working on something very much in this vein right now actually. I have a super high-dim problem where due to sparsity the problem becomes low rank during sampling. Wanting to then switch from doing inverses via Cholesky to inverses with Woodbury -- but quite fiddly numerically so far...

oh thanks! my role was tiny though, the first author (a master's student (!)) is fantastic and did the work

At a conference this summer a guy referred to papers from the late 90s as «turn of the last century»…

Hutch++: Optimal Stochastic Trace Estimation We study the problem of estimating the trace of a matrix $A$ that can only be accessed through matrix-vector multiplication. We introduce a new randomized algorithm, Hutch++, which computes a $(1 \pm ...

Been a while since I looked into this but the Hutch++ paper has some optimality results I think as well as some refs to variance reduction techniques arxiv.org/abs/2010.09649 ( from 2021 though, so I'm sure a lot has happened since )

autodiff is cool and all, but have you tried manually computing the gradient to avoid too many variables on the autodiff tape..??? (screenshot of me trying to avoid autodiffing through large Gram matrices)

Doctoral Researcher (simulation-based inference) / Väitöskirjatutkija (simulaatio-pohjainen päättely)

I'm looking for a Doctoral Researcher (PhD student) to work with me on simulation-based inference at Data Science Research Centre, Tampere University Check the link for details and send an application before October 10th.

tuni.rekrytointi.com/paikat/?o=A_...

Playing around plotting some chemical structure embeddings coming from a VAE, and colouring by local intrinsic dimension looks almost like these colour blind test images.

Éin av fire er villig til å lyge for ein tusenlapp ­

«– Når det ikkje er nokon risiko for å bli avslørt, og inga form for straff for å lyge, ventar vi frå standard økonomisk teori at alle skulle ha loge, seier Bjorvatn.»

Økonom overrasket at folk flest ikke tenker som økonomer

JOBS ALERT! PhD position & 3-y postdoc position in statistics within my research group at @ocbe.bsky.social @uio.no
Both are linked to a project funded from the Research Council of Norway on Integrative Bayesian clustering for high-dim data in omics

Deadline: 13 June 2025

More info & links below!

Associate Professor at the Oslo Centre for Biostatistics and Epidemiology (276924) | University of Oslo Job title: Associate Professor at the Oslo Centre for Biostatistics and Epidemiology (276924), Employer: University of Oslo, Deadline: Sunday, May 18, 2025

Join us in beautiful Oslo!
Associate Professor @ocbe.bsky.social
@uio.no
Deadline May 18th
www.jobbnorge.no/en/available...

British men urged to join ‘Dad strike’ calling for more paternity leave Exclusive: Fathers planning protest with babies in London on 11 June to highlight UK’s ‘rubbish’ statutory leave, least generous in Europe

Glad to see this happening! I never understood why UK dads were not marching in the street already when I saw that my UK employer only offered *two weeks* paternity leave. Coming from Norway I had to trade in a minimum 4 months leave for a measly two weeks -- time with my daughter I won't get back.

I’d also be interested in this, Martin. How it is structured, what is covered, what the pre-requisites are etc

I think that could indeed be quite illuminating. Often we approach these topics trying to make them practical without requiring too much mathematical setup. You don’t *need* stochastic process theory to get at the core of GP regression, but you might need it to understand the finer details later on

Interesting post! I helped run a similar module in Cambridge last year were we tried to cram in too much, covering basics of GPs *and* DPs. If the emphasis is on BNP I think one must mention DPs at some point, though it is a step up in abstraction compared to GPs. «distribution of distributions» etc

It can be bounded from below and above, by \sqrt{n} and n, respectively. These cases each reflect say an extreme lengthscale, ell=\infty or ell=0, and so it appears almost like a measure of complexity. I'm struggling to come up with an intuitive explanation.

Does anyone have any ideas/references?

Here, in the context of GP regression, K refers to the n x n Gram matrix associated with a set of inputs and a covariance function, while lambda_j refers to the eigenvalues of K.

Working on a model involving some GP regression, coupled with a horseshoe prior for variable selection. In an attempt at counting the number of effective parameters of my model, the following quantity pops out quite naturally. Is this a well known quantity? 1/2

Mathematics & Statistics - dScience – Centre for Computational and Data Science Read this story on the University of Oslo's website.

Maybe I'll add that the call is very general across "computational science", encompassing biology, chemistry, physics etc. alongside maths & stats. Bayesian machine learning is just one of the projects within our department, see the link below
www.uio.no/dscience/eng...

DSTrain - dScience – Centre for Computational and Data Science DSTrain is a 5-year postdoctoral programme that will award 36 postdoctoral fellowship positions of 36 months each in two calls over the programme period within the overarching frame of data science. T...

Second call for the DSTrain Postdoctoral programme @uio.no is up! The university is hiring 18 postdocs across 9 research areas and over 60 loosely defined projects.

Come work on Bayesian machine learning with us in beautiful Oslo

Deadline April 6th, please share!

Are you assuming you are able to compute the derivatives as well, or simply that you know they are non-negative?