I was invited to give this plenary talk at the Electronic Imaging Symposium this past February. The recording is now available on YouTube www.youtube.com/watch?v=s59U...

Silicon Valley crosswalk buttons have been hacked to hilarious effect

ruined by tariffs

Most people don't understand that volatility alone can kill your investments start with capital P lose 5% one week: (1-0.05)*P gain 5% the next week: (1+0.05)*(1-0.05)*P. repeat the pattern for 1yr (52 weeks): (1+0.05)^26 *(1-0.05)^26 *P = 0.88P You lost 12% of your money.

bias-variance tradeoff

It's a cruel irony that those who orchestrated this stock market crash are already insulated from its effects. They will be fine. The real tragedy is that it's the ordinary person – the retiree, the small investor, the hardworking individual – who will suffer the most.

"That Mr. Musk has come to hold so many of the same beliefs about social engineering and economic planning as his grandfather is a testament to his profound lack of political imagination, to the tenacity of technocracy and to the hubris of Silicon Valley." www.nytimes.com/2025/04/04/o...

Every profession has its personality types. In my experience the most difficult people to work with are cynics - no matter how talented they may be. Healthy skepticism is positive and curious. Cynicism is not - it's judgmental. The difference is under-appreciated. 1/2

Autoregressive models have become more popular recently. Speaking to people, I get the sense most folks are unaware of the history of the topic in signal processing. In particular, AR across scale got a lot of attention in the 90s. There are a lot of great papers on the subject but one stand out…..

A cipher (or cypher) is an algorithm for performing encryption or decryption - the foundation of all of cryptography research and practice. Not widely appreciated is that the word cipher has its direct origin in the arabic word صفر (sifr) meaning zero. Andصفر itself roots to sanskrit शून्य (śūnya)

give this kid a PhD

A man walks into a library and asks whether they have a book on Pavlov’s dog and Schrodinger’s cat. The librarian thinks for a moment and says “it rings a bell, but I’m not sure if it’s in or not”

I posted this thread four months ago. To my knowledge not a single one of these 10 basic features has been implemented on this website. It can’t be that hard.

Model Distillation

In some ways it's harder than ever to land a research job these days - both in academia and in industry. As you prepare for a job talk, consider giving it the following structure - it's hard to give a bad talk that's built like this.: 0/n

The Power of Context: How Multimodality Improves Image Super-Resolution Kangfu Mei, Hossein Talebi, Mojtaba Ardakani, Vishal M. Patel @docmilanfar.bsky.social Mauricio Delbracio tl;dr: condition SR model on RGB+depth+segmentation (which you can predict from RGB)-> PROFIT arxiv.org/abs/2503.14503

Tweedie's formula is super important in diffusion models & is also one of the cornerstones of empirical Bayes methods. Given how easy it is to derive, it's surprising how recently it was discovered ('50s). It was published a while later when Tweedie wrote Stein about it 1/n

Tweedie's formula is super important in diffusion models & is also one of the cornerstones of empirical Bayes methods. Given how easy it is to derive, it's surprising how recently it was discovered ('50s). It was published a while later when Tweedie wrote Stein about it 1/n

Don’t be fooled - SakanaAILabs has very likely already submitted dozens of manuscripts to various conferences to test out its model before productionizing in this sham “experiment”. It’s not an experiment, it’s arbitrage - a core expertise of David Ha from his finance days 1/n

1st one that we know of

really, how can you

The tech industry in Silicon Valley runs on the considerable talents and work ethic of folks who were not born here. Many of these folks were also trained in this country in our Universities. This is our strength. If we mess with this virtuous cycle, it is at our own peril.

In the 50s a sophisticated theory of how to look for things was developed. Rooted in operations research, it was largely motivated by the loss of an H-bomb and a nuclear submarine in the middle of the ocean Closely related to concepts in reinforcement learning 1/2

With today's Turing, it's worth highlighting Barto & Sutton's view of "related work": "In this book, we consider all of the work in optimal control also to be, in a sense, work in reinforcement learning" It's an example of reframing & renaming in CS that's somehow both myopic & maximalist. 1/2

Breaking: Faced with a strong economy, peaceful neighbors, and loyal allies, the United States has tapped into its strategic wacky reserve to boost world mayhem.

Have you said thank you once to Tweedie?

wrecking the economy is going great

Einstein’s eulogy of the great Emmy Noether (1882-1935) “In the judgment of the most competent living mathematicians, Fraulein Noether was the most significant creative mathematical genius thus far produced since the higher education of women began”

meet Chris J Li - this titan of thought has single-handedly conquered the fields of machine learning, optimization, statistics, reinforcement learning, and federated learning. he's not the visionary we want, but judging by the current state of affairs, he may be the one we deserve

Reinforcement Learning

When x and y are independent random variables, their joint cumulative dist function (CDF) is the product of the individual CDFs: F(x, y) - F(x)*F(y) = 0 If x,y are only uncorrelated, then Cov(x,y) =0. Can we express uncorrelated-ness in terms of the CDFs as well? Yes. 1/2

excellent use of negative space

Not one VC would ever fund a startup to do the kind of hardcore optimization work that DeepSeek did. Every VC firm should be asking themselves why.

California housing is not expensive because we are running out of land. You can get a pretty cheap house in the Central Valley. California housing is expensive because we use zoning to cap the number of homes that can be built in the places with jobs and economic opportunity.

Prof. on website vs Prof. in person

Theorem: The more advanced the math, the thinner the books Corollary: In the limit there is no text for the most advanced topics

The Laplacian operator can be expressed as the difference of a pair of smoothing operators. Here’s a canonical demonstration

best remote

rookie mistake

“We have met the enemy and he is us”

watching dudes pull all-nighters before a conference deadline

Picasso's le taureau study is a masterful illustration of distilling an idea to its core essence. A deliberate and careful stripping away of details - especially when exposing an idea in a paper or lecture - can lead to clarity, a more powerful message, and impactful insights

If you’re an AI startups, or interviewing w/ one ask: What are you the best in the world at? Do you offer a service, formula, or delivery method you invented? Is there something you do that’s patentable or a unique user experience? Have you identified and isolated a market segment? If not, walk

the five horsemen of the apocalypse

the reviewer’s insightful suggestions have made our paper more accessible

academic social media

2025 is … a perfect square 45 x 45 = 2025.   a sum of cubes: 1³ + 2³ + 3³ + ... + 9³ = 2025 a perfect square when you add 1 to each digit: 3136 = 56² a perfect square when you increase the first digit by 1: 3025 = 55² Maybe the only number with all these properties

Hard to understand what to make of this. It shouldn’t be possible to co-author some 150 papers a year (in 2024 alone) and have any idea what’s in them. Past a certain point it seems meaningless to keep track, and downright bizarre to brag about.

that tracks