“I have this absurdly over-simplified model of reality, and I’m concerned about my choice of prior distribution.”

Scoop w/ @NicoAGrant & Dina Bass: Google AI research manager Samy Bengio has resigned from his position, following the ousting of Ethical AI leaders @timnitGebru (@dair-community.social/bsky.social) and MMitchell bloomberg.com/news/articles/…

Today we achieved the dream. Made a function called `missyElliott` that takes a data structure, flips it, and reverses it. I have ascended.

Israel's Iron Dome defence system and rockets launched from Beit Lahia in the Gaza Strip rise into the night sky on May 14 📸 Anas Baba

Read this review of an ML paper at an IEEE conference. Reminded me of the quote from Wendy Rhoades in Billions: “It's like academia, everything's a big fucking deal because the stakes are so small.”

We have at least 5 postdoctoral positions at MIT CSAIL for first-generation college students, individuals underrepresented in graduate education at MIT, and others overcoming significant challenges in path toward graduate school -- applications due June 24th

Diffusion models are another type of generative models, besides GAN, VAE, and flow models. The idea is quite smart and clean. It is flexible enough to model any complex distribution while remains tractable to evaluate the distribution. lilianweng.github.io/lil-log/2021/0…

Life hack: eat cookies while reading difficult papers and you'll understand them 20% better

“These models are really castles in the air. They have no foundation whatsoever.” -J. Malik

slides for my course next week getting necessary memes now

Tensor + edge detect + object recognition + content aware full = Magic eraser on Pixel 6 👀

can It's only 79 or 2334 papers... my bigger concern is that I don't believe there are 2334 publishable results in Machine Learning this year.

I'm seeing this a lot in my timeline. Strongly disagree with "research should be X" and "if your research has a Y, you're doing it wrong". Sometimes there's a path, sometimes it's all weeds. Sometimes it's play, sometimes not... so what? As Atyiah said, the goal is to understand

I have shouted this very thing...

Google is so aggressive about mining e-mail receipts for data to sell ads that Amazon no longer tells you what you bought in their emails, and now I honestly can’t tell which company is being the bigger asshole to customers.

Hell yeah! One of the greatest wuxia movies ever made is getting a sequel!

being a statistician is fun cause you can tell people that the data they worked so hard to collect isn’t enough or won’t answer the questions they want answers to and move on while everybody hates you

"'I have this math competition experience, that ... you have to be fast,' [Wang] said. 'But June is the opposite. … If you talk to him for five minutes about some calculus problem, you’d think this guy wouldn’t pass a qualifying exam. He’s very slow.'" quantamagazine.org/june-huh-high-…

New Fields medalist June Huh recalls past conversations with older Fields medalist Heisuke Hironaka. I wonder if this is what the Turing Test feels like from a language model's perspective

The success of diffusion models is a great example of the impact that a new mathematical paradigm can have. All the hyper-parameter tuning in the world can't beat a few lines of thoughtful math.