The idea that every social problem needs a technical solution is so myopic. If you're worried about the smarts of the population, why not just fund public education? And if you're worried about Skynet, maybe try regulating the AI industries...
If you are planning to submit a paper to @neuripsconf.bsky.social and you need to create a new @openreview.bsky.social profile, do it today:
OpenReview's profile moderation process is stricter than before, thus it may take longer for new users to have their account activated (up to 2 weeks).
We know models aren’t great at counting the “r”s in “strawberry” but why? The answer isn’t as simple as tokenizers!
I wrote and posted this to Instagram a couple of days ago and it became my most-read poem ever. I’m honored I get to feel big feelings alongside you all. I’m posting it here too and want to use this space more consistently. Hello friends ❤️
The only reasonable response from Americans is to protest in the streets of every city until Congress removes the President.
Remarkable that the US is bombing a country of tens of millions of people and crashing the world energy market, but the American mainstream media is captivated by a pilot rescue story.
Some reports say over 500 schools, 55 libraries, & 25 universities hit.
You can debate the numbers, but hitting Sharif University & Beheshti is like hitting MIT & Stanford. I keep wondering: How would the scientific community respond differently if it was those universities? What’s the difference?
Absolutely heartbreaking. So many mathematicians I know studied there, men AND women.
Maryam Mirzakhani studied there.
If this isn't enough to cancel/relocate from the US the International Congress of Mathematicians #ICM2026, nothing will be.
It is insane that Americans will wake up and go to work tomorrow.
We should be on strike until the president is impeached, convicted, and tried for war crimes.
Runtimes by batch size, hidden layer size, and parallel/serial computation
Our new preprint on parallelizing training of temporally precise spiking neural networks is out!
We show up to 44x speedups over a conventional sequential baseline. 1/N
Image with text to email your senators to urge them not to confirm Jim O’Neill as Director of the National Science Foundation.
Email your senators to urge them not to confirm Jim O’Neill as Director of the National Science Foundation.
He's not a scientist, has no relevant expertise, promoted vaccine misinformation, helped gut the CDC, and his Silicon Valley ties raise questions about protecting the public interest.
(1/2🧵)
Woah you have vending machines with kurkure??
Obviously systematic change trumps all…but if you are going to make a personal choice, turns out the best thing you can do is get rid of your car.
Straight of Hormuz, with Buddha imposed, text: The realization that the lack of oil isn't causing your suffering, but the desire for oil itself, on your path to enlightenment.
Possible solution to the Strait of Hormuz?
What they actually argue (see their last section) is that *if* a neural network did these things, then it must be implementing a symbolic system under the hood, but this would shed no light on cognitive architecture itself (i.e., the algorithms that are being implemented).
This gets to the heart of modern debates: are neural networks fundamentally alternative hypotheses about the nature of intelligence, or are they particularly powerful/efficient implementations of classical ideas about intelligence? Are certain cognitive features in some sense inevitable?
By which I mean that they let me do a lot of non critical tasks like say, making UIs, but they don’t radically transform my core tasks yet
The Cowley Group at CSHL has an opening for a bioAI PhD student to start Fall 2026 to work on closed-loop AI models for visual processing (see below). You *must* have a Master's degree in a quant/eng/cs field.
www.cshl.edu/phd-program/...
Please reach out to me if interested!
I'll be at Cosyne.
DNN models of the brain are getting bigger. Are we replacing one complicated system in vivo with another in silico?
In new work, we seek the *smallest* DNN models of visual cortex, balancing prediction with parsimony.
It turns out these compact models are surprisingly small!
rdcu.be/e5H8G
What arxiv.org/abs/2512.24873
google doesn't endorse marriage through differentiation for constants and exponentials
"I'm a mathematical constant and my partner is an exponential function she insists we get differentiated but I am worried should I agree?"
I am here to torch every soft-focus retrospective on statesman and institutionalist Robin Vos. No one has done more to break the state of Wisconsin than this man.
Here's a version that lets me look at individual layers and heads (a layer just being the mean of all of its heads). Notice some heads really care about specific tokens that others don't.
You wrote “average attention score as the prompt progresses” so I was just clarifying for myself average over what
So the opacity is a running average score?
What do LLMs see?
I wrote a lil' tool that extracts the attention matrices out of open models and creates this typing visual, with each token's opacity changing according to its average attention score as the prompt progresses. Dimmer words are considered less important to the model.
Screenshot of the vancouver metro area rendered by the Stamen Design water color tileset, and it's gorgeous. it looks like a real water color painting
i just discovered Stamen's Watercolor tileset and it's just beautiful! I wish all maps where this beautiful.
Bring more whimsy and beauty to websites
We had this beautiful internet, and now it's filled with slop. Outside of my circles, I don't trust digital interactions with people, and I know they don't trust me either. (Some effects spill over into real life too.)
Was it worth it? Time will tell.
