Joshua Grochow

@joshuagrochow

Research: TheoryCompSci, pure math, complex systems Other: climate; covid; equity, inclusion, & accessibility Assoc. Prof. @ CU Boulder Comp. Sci. & Math Views my own @joshuagrochow@mathstodon.xyz (& prev twitter)

516 Siguiendo2515 Seguidores1745 Publicaciones

Bluesky

¿Es tu perfil?Reclámalo y joshuagrochow.dirty.social es tuyo.Reclamar perfil

Joshua Grochow@joshuagrochow· Jul 12

I think this same thing applies to math and (T)CS - as with much of the rest of the essay. (Literally change the one word "philosophy" to "mathematics")

But when I think about the concrete experience of reading, writing, and doing philosophy, I am reminded of what it is concretely to strive to live up to one another. I think of the distinct feeling of being in the thrall of another person’s mind and the rush of gratitude that they had these thoughts and wrote these words, for me now to be able to take them in. These feelings of awe and appreciation are inextricably linked to the human recognition of what it must have taken for someone to write this: what they must have done, the time they must have spent, the strain and struggle, the activities foregone, the desires deferred, the times they nearly quit, the times they did quit and came back and started again. How astonishing we really are; and how exhilarating it is to be able to recognize each other in this way.

Joshua Grochow@joshuagrochow· Jun 27

I just read this for the first time and it was so good! Worth every word @kyliu99.bsky.social wrote and agonized over, and worth every second it took me to read the 8380 words.

[excerpt from the novellette]

I’m arrogant enough to demand the Reader’s complete, total attention. The Reader should likewise demand that the Writer puts all of herself into her text, to infuse it with her soul. We deserve each other. It’s an act of love to read and an act of love to write. However idealized or pretentious that may sound, it’s also the truth.

Joshua Grochow@joshuagrochow· Jun 9

Scene from Star Trek: The Next Generation

Data [looking at a drink he is holding]: I hate this it is revolting!
Guinan (the bartender): More?
Data [smiling joyously]: Please.

Joshua Grochow@joshuagrochow· Apr 21

. @arula-ratnakar.bsky.social you might be interested in next week's IAMP one world math physics seminar: Trans-series in quantum field theory www.iamp.org/page.php?pag...

Upcoming Seminars

April 28, 2026
Marcos Marino (Geneva)

Trans-series in quantum field theory

Link will be available at https://researchseminars.org/talk/IAMP_seminars/189/

Abstract:
In quantum field theory (QFT), perturbative series give only asymptotic approximations to observables. Is it possible to upgrade these series to exact results? We know that this can be done in some situations -like the one-dimensional WKB method- by promoting perturbative series to so-called trans-series. Trans-series are generalizations of perturbative series that include exponentially small non-perturbative effects. In this talk I will provide an overview of trans-series in QFT, where these non-perturbative effects are typically due to instantons and renormalons/condensates. I will illustrate their general properties with concrete examples in asymptotically free theories in two dimensions, and I will give evidence that, in integrable field theories, some observables can be exactly captured by the resummation of trans-series.

Joshua Grochow@joshuagrochow· Nov 26

When I search bsky for "MIT study finds AI" here's what I get:

Bsky posts

From CNBC: MIT study finds AI can already replace 11.7% of U.S. workforce

From OG @oliviagoddess.bsky.social: MIT study finds 95% of businesses using ai aren't making a positive return on the investment (ie they lose $$$)

From Hacker News Best @hnbest.bsky.social: MIT study finds AI use reprograms the brain, leading to cognitive decline

From AI Firehose @ai-firehose.column.social: An MIT study finds that while AI boosts accuracy in misinforamation detection by 21%, it simultaneously lowers long-term discernment skills by 15.3% over four weeks, stressing the need for AI systems that foster critical thinking

From Grzegorz Wapinski @grzewap.bsky.social: MIT study finds that AI doesn't, in fact, have values | TechCrunch

From Renaud Lifschitz @nono2357.bsky.social: AI still doesn't understand the word 'no', MIT study finds

Joshua Grochow@joshuagrochow· Nov 26

"MIT study finds AI can already replace 11.7% of U.S. workforce" (CNBC article). That's not what the study did. What it did was *assume* that AI can replace certain jobs *because commercially available products are being used for tasks from those jobs*. From the study: iceberg.mit.edu/report.pdf

AI System Capabilities: We catalog over 13,000 production-ready AI tools from Model Context Protocol implementations (software development tools), the Zapier automation platform (workflow systems), and the OpenTools directory (specialized applications). These represent AI capabilities that can be packaged into deployable systems for specific occupational contexts, rather than raw frontier model performance on academic benchmarks. To align these tools with the Bureau of Labor Statistics (BLS) skill taxonomy, we develop a semi-automated mapping pipeline. Specifically, we use in-context learning with large language models to infer which skills each tool can perform, based on its task descriptions and metadata. For calibration, we split 600 occupations as training prompts and reserve 300 occupations for validation, enabling the model to learn skill-task correspondences from human-labeled BLS tasks. The model then predicts skill coverage for the tool set, which is manually reviewed to ensure consistency. This hybrid approach allows us to generate skill capability profiles for AI tools at scale, while retaining human oversight to correct systematic errors and validate uncertain cases. The result is a skill-level capability matrix that enables direct comparison between human job requirements and AI system functionality across the same dimensions.

Joshua Grochow@joshuagrochow· Nov 5

Is this...good?

You scored 100%

Stop Hate UK
Stop Hate. Start Here
A joint project

Joshua Grochow@joshuagrochow· May 30

StackOverflow may be dying (genAI is pretty decent at the basic mechanics of coding), but the Theoretical CS StackExchange (while always a smaller community) is still basically just chugging along:

Chart of new questions per month on the TCS StackExchange, 2016-2025. Wiggly line like a stock price, maxes around 80, min around 30. Slight downward trend but to the naked eye the average looks to be around 45-50, pretty consistently since 2019.

Joshua Grochow@joshuagrochow· Dec 6

From the article. I'll admit I was skeptical just based on the quoted post, but I (think) I like where the article ended up.

Something similar has to happen with academic integrity in a world where LLMs are now ubiquitous. We need to think about academic integrity as a bigger concept rooted in educational values, values that are tied to student engagement, effort and learning.

I’m convinced we’re significantly underestimating the degree and kinds of changes that need to happen in educational institutions to deal with the existence of generative AI technology. These changes need to deal not only with the technological capabilities, but also with the deterministic way the technology is being framed by those who are developing and boosting it.

Some of this boosting is happening inside of higher education institutions that have decided—without a ton of hard evidence, by the way—that AI is an inevitable part of our collective and individual futures. I have no desire to wall education off from artificial intelligence, but the notion of its inevitability is something I think we should resist with what remains of our might.

Joshua Grochow@joshuagrochow· Nov 9

18. @pwnallthethings.bsky.social Any finite subgroup of n x n complex matrices has an Abelian normal subgroup of index bounded by a function of n alone (and by (n+1)! for n \geq 71). Collins (J. reine '08) also has an analogue over arbitrary fields, but the statement is a little more complicated:

$Theorem A. Let k be an algebraically closed field of positive characteristic ell and let G be a finite subgroup of GL(n,k). Put G-bar = G/O_ell(G). If n \geq 71, then G-bar has normal subgroups N,L such that (i) N is abelian (ii) L = E_ell(G-bar), and (iii) [G : LN] \leq (n+2)! if ell divides n+2, and (n+1)! otherwise If the bound is achieved, then E(G-bar) is simple and G-bar/Z(G-bar) is isomorphic to S_{n+2} or S_{n+1}, respectively. Furthermore, O_ell(G)=1 unless n is congruent to -1 (mod ell), in which case O_ell(G) may be elementary abelian.$

Joshua Grochow@joshuagrochow· Nov 8

14. @chantzy.bsky.social Sometimes things look discontinuous because we don't have all the information. For example, in the attached slide the 2D shadows appear to have "corners", but w/ all the information (all 3 parameters) we see only smooth surfaces. (my slide, I'll have to dig up image credit)

Hidden variables create (the appearance of) complexity.

Three sequences of 3D figures and their 2D projections/shadows:
- A rotating donut. When it becomes vertical enough, corners appear in its 2D projection
- A smooth surface bending a bit creates a football-like shadow
- A smooth surface bending in a different way creates an X-like shadow (called a swallowtail)
- A smooth surface bending in yet a third way creates the appearance of corners, that expand to two meeting circles, that eventually separate.

Joshua Grochow@joshuagrochow· Nov 7

I wanted to share some things I think are amazing and cool. Each like -> one theorem (or definition or phenomenon or example or...). #MathSky

Poorly edited text graphic reading "Get to know my math taste. For every like, I will post one theorem I love ❤️"

Joshua Grochow@joshuagrochow· Oct 25

New (*draft) slide just dropped, that I want to include at the start of all my online talks. Thoughts or feedback? #MathSky #AcademicSky

Slide title: Thank You for Hosting Remote Talks

Body of slide:
Many researchers cannot travel because of:
- Visa issues
- Disability
- Personal or family health issues
- Money - personal, university, country
- Family responsibilities
- Work responsibilities
- (Climate change & environmental cost)

[Highlighted] Remote events help tie us to the community!

Joshua Grochow@joshuagrochow· Sep 18

It has become unbearable to manually scroll back and forth between the main text and the references at the end of a paper. #AcademicSky #MathSky #SciSky 🧪

$Bernie Sanders meme "I am once again asking you...if you use LaTeX, \usepackage{hyperref}"$

Joshua Grochow@joshuagrochow· Jul 10

Phoebe teaching Joey French, from Friends meme

Phoebe: Step 1
Joey: Step 1
Phoebe: Step 2
Joey: Step 2
Phoebe: Step 3
Joey: Step 3
Phoebe: So the answer is... [looking expectantly]
Joey: [looking excited, like he figured it out] So what's the answer?

Joshua Grochow@joshuagrochow· Jun 28

Tag yourself. #Algebra #Math 🧮 #physics

Alignment chart:

Lawful Good: non-commutative (hyphenated)
Neutral Good: noncommutative (not hyphenated)
Chaotic Good: non commutative (with a space)

Lawful Neutral: non-Abelian (hyphenated, capital A)
True Neutral: nonabelian (not hyphenated, lowercase a)
Chaotic Neutral: non Abelian (space, capital A)

Lawful Evil: non-abelian (hyphenated, lowercase a)
Neutral Evil: non abelian (space, lowercase a)
Chaotic Evil: nonAbelian (no space, no hyphen, capital A)

Joshua Grochow@joshuagrochow· Apr 1

Best April Fools I've seen so far is the website from the Lab for Atmospheric & Space Physics at CU Boulder: lasp.colorado.edu It's worth clicking & scrolling, they really went all out! (h/t Liz Bradley) 🧪

Joshua Grochow@joshuagrochow· Feb 16

The 19 April 2024 meeting of Functor Categories for Groups will be hybrid! www.lancaster.ac.uk/maths/fcg/ To register for the event or to receive the talk links, please email the organiser Nadia Mazza www.lancaster.ac.uk/maths/people... #MathSky

Groups come in many fascinating forms and shapes. In this meeting, we will survey a selection of original techniques in the study of some infinite groups, such as right-angled Artin and Coxeter groups, profinite groups and groups associated to a certain Cantor action. Techniques may involve graphs, the use of micro-supported actions, or Mackey functor methods.

Provisional timetable
13:15-14:15 Jakub Waniek (Lancaster)

14:20-15:20 Dominik Francoeur (Newcastle)

15:20-15:50 refreshments

15:50-16:50 Gemma Crowe (Heriot Watt)

Titles & abstracts nearer the time.

Joshua Grochow@joshuagrochow· Feb 15

*Hybrid* Summer School on Formal Techniques, May 25 - June 3, 2024 fm.csl.sri.com/SSFT24 (US Pacific Time) Applications due April 30th. I'm particularly excited to see J. Urban's "Combining #MachineLearning and Theorem Proving" and L. de Moura & D. T. Christiansen's "#Lean 4"

The lecturers at the school include:

Josef Urban - Combining Machine Learning and Theorem Proving
Marsha Chechik - Elicitation and Formal Reasoning about Normative Requirements
Leonardo de Moura and David Thrane Christiansen - The Lean 4 programming language and theorem prover
Cesare Tinelli - Modeling and analyzing reactive systems with logic-based symbolic model checkers
Armando Solar-Lezama - Neurosymbolic Programming for Better Learning

Joshua Grochow@joshuagrochow· Feb 13

SIGEcom Winter Meeting this Thursday (15 Feb 2024) online: www.sigecom.org/meetings/win.... Free registration This year's topic is behavior models with a great lineup of speakers!

All times listed in US Eastern Time Zone (ET). See here for detailed schedule including abstracts.

11:10 - 1:00pm: Introductory Talks (Seminar room)

11:10am - 12:00pm: Jon Kleinberg - Behavioral Agents in Algorithmic Environments
12:00pm - 12:10pm: Break
12:10pm - 1:00pm: Ori Heffetz - EBRD and Market Design

1:00pm - 2:30pm: Social Break

1:15pm - 1:45pm: Student Fireside Chat with Noam Nisan and Al Roth (Lounge)
2:00pm - 2:30pm: SIGecom Advisors Meet and Greet (Gazebo)
2:00pm - 2:30pm: SIGecom Execs Office Hours (Garden)

2:30pm - 5:00pm: Spotlight talks (Seminar room)

2:30pm - 3:00pm: Modibo Camara - Computationally Tractable Choice
3:00pm - 3:30pm: Ryan Oprea - What Makes a Rule Complex
3:30pm - 3:45pm: Break
3:45pm - 4:15pm: Gali Noti - Learning When to Advise Human Decision Makers
4:15pm - 4:45pm: Nicole Immorlica - Data, the Fundamental Particle of Interaction

4:45pm - 5:00pm: Concluding Remarks (Seminar Room)

5:00pm - 6:00pm: Closing Reception (Garden)

Joshua Grochow@joshuagrochow· Oct 25

#AcademicSky #MathSky

Bernie Sanders w/ text overlaid:

I am once again asking departments

To please say what university, department, & position a rec letter is for when requesting letters by email.

Joshua Grochow@joshuagrochow· Oct 11

(Some people tried to make "skeet" happen, but thankfully I see that falling by the wayside)

Picture of Rachel McAdams in "Mean Girls", wearing a Santa hat, with white text overlayed: Stop trying to make "skeet" happen

Joshua Grochow@joshuagrochow· Sep 27

Convenient covid air decay calculator 👇 based on UV index, temperature, and humidity. Seems esp. useful for places with poor ventilation. Website includes background and caveats of the model.