David's personal blog.
Typesetting tables in latex is annoying. In practice, I usually have the tabular data in some programming-friendly format and need to put it in latex. I forget how to do this nicely, but it's extremely simple using the tabulate package in python. import tabulate from itertools import batched # Data A = [ 1 , 2 , 3 , 4 , 5 , 6 , 7 , 10 , 11 , 14 , 15 , 17 , 19 , 21 , 23 , 26 , 31 , 34 , 35 , 39 ,…
There is no reason why amateurs can't make worthwhile contributions to research mathematics. But there is also no reason why every mathematician must listen to any particular amateur. 1 1 This is much more true now that it's easier to write mathlike documents with help from an LLM. I don't have time to offer personal advice or guidance to everyone who contacts me. 2 2 If I have sent you to this…
I write most professional correspondence in either plaintext (markdown) or latex+pdf. I recently had to make two versions of the same document in latex and Google Docs. 1 1 This is unnatural. The medium of communication affects both the reading and writing experience. But I don't explain why , and instead merely comment on how . Eventually the latex+pdf version will be made public, so I write the…
I've heard it said that reading LLM prose is a bit like remembering a dream. It can be ok if you were the one who had the dream, but all details are fuzzy and unhelpful if you try to describe it to others. Nonetheless, I think it can be useful to share LLM-generated or LLM-assisted writing with explicit AI content warnings . For now, I will give AI content warnings in approximately the following…
You know what's great? Writing things down and sharing them. I want to do that more here. To make this happen more, I'm going to intentionally, deliberately lower the bar for what I think is suitable to put here. I might get lucky and have a quantity-breeds-quality aspect. But I also know that I've been emailing and zuliping a manner of notes that I've written over the last few years that aren't…
At today's Paper Talk session cohosted by IDI, I'm giving a talk today on the paper "The Free Transformer" by Fleuret ( preprint available here ). This paper describes a modification of the standard decoder-only transformer that learns to condition token generation on random latent variables learned without supervision. We describe some of the intuition and background for the new architecture. My…
This post is larger than 10000 bytes, which is above the limit for this RSS feed. Perhaps it is long or has embedded images or code. Please view it directly at the url.
I have left academia and accepted a job as a Principal Engineer at the Institutional Data Initiative at Harvard Law. By "principal engineer", think "principal software engineer", "principal data engineer", and "principal data scientist". I appreciate the academic freedom I enjoyed at ICERM and the University of Warwick. One enormous compliment I have for the Academy is that the flexibility to…
This post is larger than 10000 bytes, which is above the limit for this RSS feed. Perhaps it is long or has embedded images or code. Please view it directly at the url.
Next week, code4math will be officially launching as an AIM Research Community. The opening will come in 4 sessions of talks spread across two days: September 8 and 15. On each day, there will be a morning session and an afternoon session (relative to where I am in US Eastern), roughly to be accommodating to interested participants in many different timezones. We have a blurb that we wrote and…