Leo Du

Positive semi-nondeterministic PhD student @jhuclsp currently visiting #rycolab ETH | previously @uwcse | math junkie | profile pic reads "the 8th Busy Beaver"

[Leo Du @leoduw]

Love this distinction between “compiler optimization” and “programming model”. Coming back to a common example, part of Rust’s success is its programming model.

Also, great talk!

[Horace He @cHHillee]

I gave a talk at Jane Street about building ML systems!

Although it's nominally about PyTorch, the talk primarily focuses on the notion of a "programming model", and how compiler optimizations are not a programming model.

(1/4)

[Leo Du @leoduw]

- What do you call someone who trains LLMs on Apple Silicon in Metal instead of CUDA?
- FullMetal Alchemist

[Leo Du @leoduw]

@jxmnop Not to mention Rotary Embedding

[Leo Du @leoduw]

@BlackHC @keenanisalive I guess the confusion arises from the fact that the self-transition probabilities are not directly highlighted in the animation, giving the illusion that x and y might be events and not states.

[Leo Du @leoduw]

@universeinanegg @rlmcelreath It’s honestly a great idea to teach probabilistic programming in any intro stat class, Bayesian or not!

[Emily Riehl @emilyriehl]

Dominic Verity, Mario Carneiro and I just announced a new project to formalize some aspects of ∞-category theory in #Lean via the notion of an ∞-cosmos.

[Leo Du @leoduw]

@TonyTheLion2500 Stealing a line from Paul Halmos that partially applies here

[Preetum Nakkiran @PreetumNakkiran]

all theorems are correct, but some are useful

[Leo Du @leoduw]

@m2saxon @sarvk @LrzNeedResearch @aclmeeting Now imagine a band named “neurolinguistic programming” performing at Austin city limits

[Leo Du @leoduw]

@mblondel_ml @n0mad_0 @qberthet It would be incredible for the progress of many scientific fields if we have automatic assistants for these tasks!

[Leo Du @leoduw]

@m2saxon What’s worse is that there seems to be no legislation plan attempting to address this kind of problem in any way by either party.

[Leo Du @leoduw]

@TonyTheLion2500 Density of the sum of two random variables.

[Justin T Chiu @justintchiu]

wrote a short blog post on the latent variable model behind STaR: https://t.co/Wcfu8gyxKF

[Jordan Ellenberg @JSEllenberg]

I knew a lot of mathematicians were getting into formalization but I had no idea it was this much of a problem

[Leo Du @leoduw]

@taz_chu When tensors start to have more than 3 indices, this will be very handy

[Dan Roy @roydanroy]

So creative. https://t.co/tAvdZIOe4t

[Leo Du @leoduw]

"Let no one ignorant of geometry enter."

Finally, a compute can enter Plato's academy.

[Google DeepMind @GoogleDeepMind]

Introducing AlphaGeometry: an AI system that solves Olympiad geometry problems at a level approaching a human gold-medalist. 📐

It was trained solely on synthetic data and marks a breakthrough for AI in mathematical reasoning. 🧵 https://t.co/g3RFSoWNPP

[Leo Du @leoduw]

You'd think that classes in the Borel hierarchy would be named in the same language...

[Justin T Chiu @justintchiu]

wrote a short note on using parallel scans for backprop: https://t.co/QpEfGFgkSj

turns out there was already a paper on this too! https://t.co/9wIAQSI1nk

[Leo Du @leoduw]

@jonasgeiping @deliprao @tomgoldsteincs Maybe I’m in the minority but I like the ICML version (and typesetting) better!