Zachary Charles (@MatharyCharles) header
Zachary Charles (@MatharyCharles) profile picture

distributed machine learning @ google | sometimes mathematician

Seattle
Joined September 2012
There could not be a more apt metaphor for the central tension in the math community these days than seeing this on a wordpress site of all things
1
167
I don't think you read his note closely enough - he says explicitly that his own ideas may have made things worse.
1
4
228
At least some of these OAI math results build off others from the same batch. Cool to see that their models are already seeing benefits from having this enormous database of new results!
1
165
This rules - experimentation in how to communicate math and computer science is vastly underexplored. This reminds me of one the most interesting (and successful IMO) experiments in math thesis writing I've seen. theliberatedmathematician.co…
Hand painted artisanal science posters should be the new craze.
1
6
881
Here's part of Melanie Wood's thoughts, who I think does a good job of articulating something on my mind: counter-examples are actually typically easier to verify than positive proofs. Does this help explain why we've seen more progress on counter-examples?
1
95
The "Remarks on the disproof of the unit distance conjecture" has aged incredibly well. Here's Jacob Tsimerman, basically predicting something like this.
hello there the jacobian conjecture is false thanx to my close friend akhil for asking about it and my other close friend fable for working during the world cup final ((1+xy)^3 z + y^2 (1+xy) (4+3xy), y + 3 x (1+xy)^2 z + 3 x y^2 (4+3xy), 2 x - 3 x^2 y - x^3 z): \C^3\to \C^3, has jacobian determinant -2, and sends (0, 0, -1/4), (1, -3/2, 13/2), and (-1, 3/2, 13/2) to (-1/4, 0, 0)
2
8
741
2-for-1 special. Run MAGMA, get research.
Shoutout to mathematical counterexamples with zero elegance. Gotta be one of my favorite genders
2
1,045
Replying to @StefanFSchubert
Don't make me tap the sign
6
490
Not sure what you're suggesting - what does "better triage" look like in the face of the recent deluge of submissions? Worth noting that the October change was not a policy change, just more consistent enforcement. arXiv has never been a place for survey and position papers.
1
1
175
A useful meditation on LLMs but mainly I'm struck by how confident all the replies to it are. I don't think you should be as certain as you are people. E.g. see Nicolas Carlini's experiment in getting people to put 90% confidence intervals over future AI capabilities.
Given what current-gen LLMs (say, in math, but whatever) can do, I think their apparent limitations are kind of mysterious. What is the blocker preventing, at present, high quality fully autonomous work?
1
2
408
Amazing. Warren is deliberating trying to choke off investment in the construction of *new* single-family rental properties. This is profoundly regressive. Why is it OK for large investors to build and rent out apartments but not single-family homes? Many working-class Americans don't have high enough incomes or credit scores to buy a home. If we cut off investment in rental houses, then these families won't be able to live in large swathes of America: Roughly 70 percent of all residential land is zoned for single-family residences. The only way for working-class people to live in many affluent suburbs -- and access their high-performing schools -- is to rent a house. For this reason, corporate investment in single-family homes reduces socioeconomic segregation. Warren's policy effectively helps rich people keep working-class renters out of their towns -- while reducing the overall supply of housing and nudging up rents. It prioritizes populist symbolism over progressive outcomes.
1
164
I will admit - if you think wikipedia articles on math are bad, take a look at machine learning ones. Just the most shameless self-promotion. Can you guess which of these 2 edits were simply plugs for the author's niche publications?
1
1
308
I don't know when I'm going to stop Tukeyposting, but it won't be today.
1
3
383
Replying to @iamgrigorev
Very cool work. Any idea what the noise floor looks like when evaluating weight decay schedules? E.g. is a +1.5% in perplexity statistically significant if you re-ran with different initialization seeds?
1
5
1,647
Replying to @abakcus
If you remove the background, my PhD advisor had a 2 sentence paper, most of which is a Magma command.
10
1,648
Reading "Factoring Rational Polynomials over the Complexes" and this part jumped out at me. Is this "random Monte-Carlo NC" computational complexity class something well studied?
1
298
Big news - someone made duck/structural typing into a philosophy. Will wonders never cease?
1
303
I didn't realize there had been progress on the Scholz-Brauer conjecture in the last year or so! It's actually not tight. Also if you don't know what an addition chain is, I like Knuth's exposition.
1
3
3
384