Jaakko Lehtinen
@jaakkolehtinen.bsky.social
Generative AI and computer graphics at Aalto University & NVIDIA Research. @ellis.eu Fellow. https://users.aalto.fi/~lehtinj7
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.)
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.)
October 17, 2025 at 7:32 PM
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.)
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.
October 10, 2025 at 6:10 AM
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 😅
July 22, 2025 at 3:33 PM
@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..
May 8, 2025 at 7:32 AM
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.
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.
March 28, 2025 at 12:45 PM
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.
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
“if it’s smart, it’s sh*t”
is IMO even better
March 27, 2025 at 8:44 AM
“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
“if it’s smart, it’s sh*t”
is IMO even better
Great watching and learning from the stable genius 🤯
February 28, 2025 at 7:14 PM
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..
February 27, 2025 at 8:18 PM
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..
Reposted by Jaakko Lehtinen
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
February 4, 2025 at 3:16 PM
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.
It’s been a while since I last donated, but I just gave £500.
Please consider giving according to your means.
February 7, 2025 at 9:10 AM
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.
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
February 3, 2025 at 3:44 PM
Gosh I needed this amidst all the news
Reposted by Jaakko Lehtinen
📣 @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.)
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.)
January 27, 2025 at 11:38 AM
📣 @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.)
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.)
Reposted by Jaakko Lehtinen
@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-...
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-...
A free, powerful Chinese AI model just dropped — but don’t ask it about Tiananmen Square
Those who train the AI models get to decide what the truth is....
sherwood.news
January 24, 2025 at 7:39 PM
@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-...
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.
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.
arxiv.org
January 15, 2025 at 7:54 AM
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.
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 positions are on-site only — we believe in high-bandwidth F2F communication.
nvidia.wd5.myworkdayjobs.com/en-US/NVIDIA...
Research Scientist Intern, Generative AI - 2025
NVIDIA is searching for a world-class researcher in generative AI to join our research team. You will be conducting original research for generative AI applications, including the generation of langua...
nvidia.wd5.myworkdayjobs.com
January 14, 2025 at 2:16 PM
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 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.
January 14, 2025 at 2:04 PM
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
December 29, 2024 at 6:48 PM
Gutsy move pushing their own shows like this
Reposted by Jaakko Lehtinen
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
- foundations of (prob) ML
- distribution shifts
- AI-assistance with computationally rational user models
- applications in synthetic biology and biodesign
Postdoctoral researcher positions in Probabilistic Machine Learning research group, Aalto University | Aalto University
www.aalto.fi
December 23, 2024 at 1:10 PM
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
- 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 🥲
I thought it would be a lot longer until I got to teach him color constancy 🥲
December 17, 2024 at 7:36 AM
- 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 🥲
I thought it would be a lot longer until I got to teach him color constancy 🥲
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)..
Where are the equivalents of audiophiles for video?
Stills have film geeks and lovers of great lenses (like me)..
December 11, 2024 at 8:00 PM
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)..
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...
blog.neurips.cc/2024/12/10/a...
openreview.net/forum?id=bg6...
December 11, 2024 at 8:12 AM
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...
blog.neurips.cc/2024/12/10/a...
openreview.net/forum?id=bg6...
Reposted by Jaakko Lehtinen
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.
Here is a teaser of some of our recent research.
November 23, 2024 at 8:45 AM
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.
Here is a teaser of some of our recent research.
Reposted by Jaakko Lehtinen
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)
November 22, 2024 at 3:14 PM
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)