“Please let us know whether you are able to review this article[…] We would hope to receive your report by 26 December, but if you need more time we would be happy to extend this.” Yeah sure, sign me up!

Have people thought of using reinforcement learning for playing what-if games to study the potential effects of large policy changes beyond first order? Comes to mind seeing the circus associated with recent changes in the Finnish public healthcare funding

I’ve long thought the most immediate use for XR glasses would be ad blocking. ChatGPT 5.1’s take below. Maybe time for an art/social commentary project using a high-quality passthrough VR headset like Varjo’s? I bet you could distill this into a real-time model good enough for making the point.

In anticipation of seeing the new Spinal Tap next week we watched the original. It’s probably 20y+ since I saw it last, and though I loved it then, my memory still didn’t quite do it justice. A true masterpiece. Man! (Heads up: Nigel Tufnel day, 11/11, is coming up fast.)

iOS 26 is a clear design regression. UI colors appear washed out in a manner that reminds me of the disgusting early 2000s “HDR” aesthetic, and the Liquid Glass look is gimmicky. The whole thing has an air of poor taste.

@economist.com Your questionnaire on push notifications doesn’t let me tell you I don’t allow them from any of the news apps I use 😅

If you plan on skimping on LED density in your fancy indirect lighting à la mode, don’t use glossy materials directly below..

Did you know SVG and Inkscape support instancing? There's even a "Create Tiled Clones" feature in the latter. Dread the idea of drawing something like this with the raw positioning tools and copy/paste; no way to edit the appearance of the individual tiles in bulk.

“If it’s smart, it’s vulnerable” says Mikko Hyppönen — but @nurpax.bsky.social‘s version “if it’s smart, it’s sh*t” is IMO even better

Great watching and learning from the stable genius 🤯

Is Ravi & Pat’s 2001 classic on irradiance environment maps missing a factor of 1/pi from Eq. 2? Sure seems that way to me..

ELLIS Institute Finland is hiring Principal Investigators in AI + machine learning. World-class resources for research incl. LUMI supercomputer, generous starting package & professorship affiliation with a university in the world’s happiest country! Apply by March 9: ellisinstitute.fi/PI-recruit

The rabid guy with the world’s largest megaphone has Wikipedia in his sights. It’s been a while since I last donated, but I just gave £500. Please consider giving according to your means.

Gosh I needed this amidst all the news

📣 @ec.europa.eu, please force large online retailers to offer consumers choice of courier! Now I have no way to vote with my € and punish liars like UPS, who continue this charade despite having been exposed for it over and over. (Someone *was* waiting, and I have paid for home delivery. GRRRR.)

@edzitron.com on DeepSeek Good: v3 trained for $5.5m, proving that you don't need to spend half a trillion dollars on new data centers to make great models MIT licensed! Great for running on my own hardware Bad: the CCP influence is genuinely a problem for my uses sherwood.news/tech/a-free-...

Bravo 👏 An analytic theory of creativity in convolutional diffusion models by Kamb & Ganguli I’ve wondered about this very problem for two years now, with wispy intuition that points to exactly this. I never got far into experimentation, let alone theory formation. I wish I had written this paper.

Our group at NVIDIA Research Helsinki is looking for stellar research interns to work on generative AI! The positions are on-site only — we believe in high-bandwidth F2F communication. nvidia.wd5.myworkdayjobs.com/en-US/NVIDIA...

The dimension of the orthogonal group O(n) is n(n-1)/2. That’s a surprising result to me intuitively. The (dangerous as always) low-dimensional intuition would suggest they’d be much more rare.

Gutsy move pushing their own shows like this

Planning to think about what to do next over the holidays? I am hiring postdocs to my group in Helsinki, in combinations of the following: - foundations of (prob) ML - distribution shifts - AI-assistance with computationally rational user models - applications in synthetic biology and biodesign

- Daddy, why is that one spot blue? - Well, son, let me show you how the snow is actually the same color.. I thought it would be a lot longer until I got to teach him color constancy 🥲

Finally!

Seeing up close the kind of garbage a ~5mbps stream is on my superb 65” LG OLED TV has long made me think increasing the screen resolution is pointless. Sure, the edges are sharp.. Where are the equivalents of audiophiles for video? Stills have film geeks and lovers of great lenses (like me)..

Our paper on Guiding a Diffusion Model with a Bad Version of Itself was runner-up in NeurIPS 2024 best paper awards 🤯 blog.neurips.cc/2024/12/10/a... openreview.net/forum?id=bg6...

I'll be using this channel mostly to announce news from the EPFL Geometric Computing Laboratory (www.epfl.ch/labs/gcm/). Here is a teaser of some of our recent research.

My team RGL is looking for new PhD students starting 2025. If you are excited about topics like inverse rendering, compilers for graphics, and physically based modeling then please join us at EPFL. Info about the lab & admission process: rgl.epfl.ch/pages/jobs (Deadline: Dec 15)

Slang is now officially under open governance at Khronos, and what better way to celebrate than running Slang shaders directly in your browser with WebGPU? try.shader-slang.org You can also compile your shaders to a variety of targets including Metal and HLSL. I'm giddy to finally share this!