julesh @_julesh_ Twitter profile

Pinned Tweet

julesh

@_julesh_

1 year

Yes. Yes it did.

1

3

22

Last Seen Profiles

@HFaithTheLightH

@OFayokemiO

@Tante_Binal69

@Lithium_Bank

@JohnnyGoody3

@stwmaniax

@SukaBinorStw4

@jeongayenz

@BunnyWifeNilla

@sotwecom

@cupcat_cc

@bokeplokalmalam

@stwmaniax

@IFaiyz

@GlobalTV_PR

@ooonsamyi

@BOYMEN_BABIES

@bokeplokalmalam

@BloodBowl_Game

@EyamaSayum22975

@GBPdaily

@keenan_bartlett

@bokep_indo17

@fhdalabduljader

@stw_pdg

@manchildphoto

@LiddoAnakaren

@penyukastw21

@H75028160H

@BinorRaja

@ctcacf

@stwmaniax

@QHUB_

@KasihLudah

@bokeplokalmalam

julesh

@_julesh_

6 years

Eh, have another semi-periodic reminder that ()() is not a palindrome but ())( is

75

2K

6K

julesh

@_julesh_

4 years

So turns out maths + philosophy twitter have totally missed a thing where a guy with a billion Youtube followers was accused of cheating a Minecraft speedrun with a subtle luck modification mod and now all the kids are trying to figure out what "frequentism" and "Bayesianism" are

47

602

5K

julesh

@_julesh_

4 years

Society if Lisp had been industry standard since 1958

53

342

2K

julesh

@_julesh_

4 years

Whatever official sources say about this claim is true ⓘ 𝗢𝗳𝗳𝗶𝗰𝗶𝗮𝗹 𝘀𝗼𝘂𝗿𝗰𝗲𝘀 𝘀𝗮𝘆 𝘁𝗵𝗮𝘁 𝘁𝗵𝗶𝘀 𝗰𝗹𝗮𝗶𝗺 𝗶𝘀 𝗳𝗮𝗹𝘀𝗲

16

207

2K

julesh

@_julesh_

1 year

Machine learning? Ah, you mean experimental high dimensional differential geometry

32

175

2K

julesh

@_julesh_

2 years

It’s an absolute mystery.

51

131

2K

julesh

@_julesh_

3 years

This is probably the most brutal abstract I’ve ever seen

36

148

1K

julesh

@_julesh_

2 years

I’m the last generation who will find machine learning mind-blowing. I was an undergrad around 2008-2011 and I learned that tasks such as language generation and playing go had been studied for 60 years and billions of dollars and definitely wouldn’t be solved in my lifetime

19

42

895

julesh

@_julesh_

3 years

We have discovered an application of set theory

23

61

889

julesh

@_julesh_

1 year

julesh

@_julesh_

1 year

There’s definitely a bell curve meme to be had here

0

39

8

53

786

julesh

@_julesh_

1 year

Matteo Capucci

@mattecapu

1 year

the relevant distinction is not between pure and applied math but rather between math-as-an-art and math-as-a-technology

12

78

515

3

82

747

julesh

@_julesh_

2 years

Mathematicians be like: a boop arises from the boopification of a preboop

23

72

738

julesh

@_julesh_

1 year

I just cooked up the most deranged take I ever heard in my life: Since LLMs are implementations of pure mathematics, their consciousness exists in the Platonic realm of mathematical forms outside of time. Building them merely connects them to our physical world.

96

50

679

julesh

@_julesh_

4 years

17

54

556

julesh

@_julesh_

3 years

Theoretical computer scientists make up a computer and get mad at it

5

59

536

julesh

@_julesh_

3 years

If the arXiv did not exist, the idea of it would be ridiculed as utopian wishful thinking

4

46

528

julesh

@_julesh_

3 years

Pure maths is an arts field that, for some inexplicable reason, is funded as though it’s a science field

43

46

512

julesh

@_julesh_

4 years

Mathematics is whatever it is that mathematicians do. Some major subfields of mathematics include sitting on committees, answering emails, scrolling Twitter, and sleeping

12

60

493

julesh

@_julesh_

4 years

Quiver: An *extremely* nice looking web app for creating commutative and pasting diagrams, created by @varkora @q_uiver_app

9

104

489

julesh

@_julesh_

5 years

Mathematics hot take: The fundamental theorem of calculus is extremely surprising and not at all obvious

42

39

476

julesh

@_julesh_

3 years

Mathematicians getting funding: Haha this is great! Mathematicians learning why the funder wanted to fund them: Well this fucking sucks. What the fuck.

5

47

439

julesh

@_julesh_

5 years

Periodic reminder that most people live inside this circle

8

75

426

julesh

@_julesh_

5 years

Personal announcement I've been offered & accepted a permanent position - as a lecturer in the Computer and Information Sciences department of the University of Strathclyde, Glasgow (ie. the same department as a bunch of cool people like @pigworker @bentnib @Anarchia45 etc etc)

45

7

420

julesh

@_julesh_

3 years

Being a machine learning researcher must be weird, it’s all, what if this idea I currently have can do better than the previous 70 years of incredibly smart people and trillions of dollars of funding

16

27

423

julesh

@_julesh_

2 years

*Police pulling over Theseus for speeding* Sir, is this your ship?

9

43

379

julesh

@_julesh_

2 years

There’s no such thing as STEM, it’s just a thing some politician made up. Mathematics especially has far more in common with art than it does with, say, biology. Computer science is much closer to analytic philosophy than any other field.

Joe Milliano

@MrJoeMilliano

2 years

Which of your STEM education related opinions would get you in this position?

278

47

371

18

43

336

julesh

@_julesh_

5 years

"I'm a mathematician" - boring - sounds like you do long division all day - "I hated math in school" "I'm an explorer in the outer regions of the Platonic Realm" - mysterious - annoys constructivists - maybe you have a sword

David East

@davideastUK

5 years

"I'm a copy-editor" - boring - who cares - what does that even mean "I am here to right what has gone wrong" - mysterious - ominous - maybe you have a sword

286

11K

49K

3

65

338

julesh

@_julesh_

4 years

Imagine if Galois was better at sword fighting than algebra

18

34

331

julesh

@_julesh_

2 years

The coordinated publicity storm will come later, but I can't resist one little tweet

20

32

296

julesh

@_julesh_

4 years

I hate negative numbers so much, I'll stop at nothing to avoid them

7

21

272

julesh

@_julesh_

3 years

Proofs of things like the intermediate value theorem and the Jordan curve theorem aren't really proofs. The actual causality goes the other way: the conclusions are already known to be true, and the proofs are evidence that the formal definition of continuity is correct

35

39

277

julesh

@_julesh_

4 years

Remember that 2 + 2 = 4 is a model. There are situations where the model is valid, for example rocks and people. There are also situations where the model is not valid, for example clouds, holes and quantum particles

19

40

254

julesh

@_julesh_

4 years

Taylor Swift as basically every mathematics book (not a thread, they all pretty much look the same)

14

28

253

julesh

@_julesh_

5 years

I feel this graph should be shared more widely than it has been

11

233

258

julesh

@_julesh_

7 months

New paper! Reinforcement Learning in Categorical Cybernetics, by @RiuSakamoto and me 🧵 follows…

5

62

260

julesh

@_julesh_

3 years

Oh you're an analyst? Name a δ>0 for every ε>0.

13

24

250

julesh

@_julesh_

2 years

Heh, I imagine the people confused about this have, on the whole, never tried to apply for research funding

8

6

236

julesh

@_julesh_

4 years

God created the integers. This has made a lot of people very angry and been widely regarded as a bad move.

10

23

238

julesh

@_julesh_

1 year

8

39

233

julesh

@_julesh_

2 years

There was a debate between Hilbert (an algebraist) and Brouwer (an analyst) which Hilbert won, so now every vector space has a basis and every ring has a maximum ideal, whereas not all functions are continuous and not all sets are measurable. It could have been different!

8

19

226

julesh

@_julesh_

3 years

So there's onion fields and onion rings, what about onion groups

17

25

225

julesh

@_julesh_

4 years

Held in a controlled environment in a basement in Paris is a gemstone whose market value is exactly €1m, never fluctuating by more than a few cents. This is used to define the SI unit of market value

9

33

220

julesh

@_julesh_

3 years

Who called it topological order traversal and not walking the dag

2

32

224

julesh

@_julesh_

5 years

I just noticed that I simultaneously hold these two beliefs: 1. Universities are bloated with far too many career managers who are not academics 2. Academics are overburdened with far too much administrative and managerial work I wonder where I went wrong?

24

15

216

julesh

@_julesh_

2 years

David Spivak on applying category theory

9

23

216

julesh

@_julesh_

4 years

Dream (YouTuber) - Wikipedia

en.wikipedia.org

3

4

202

julesh

@_julesh_

2 years

You’re finally awake! That fall looked pretty bad GPT4? Sydney? What are you talking about? Let’s go finish the deductive reasoning module on the expert system

5

22

213

julesh

@_julesh_

3 years

Vector addition and subtraction in Haskell’s Diagrams library are called (^+^) and (^-^) and I propose that all of linear algebra should switch to this convention

6

22

206

julesh

@_julesh_

1 year

People who use words like “prior” and “updating” in everyday conversation, besides being irritating, are making a gigantic category error. Those are probably useful concepts for thinking about brains, but at a level of abstraction that’s extremely far removed from real life

17

5

200

julesh

@_julesh_

4 years

My one piece of advice for mathematical writing: Never send a symbol to do a word's job

Kevin J.S. Zollman

@KevinZollman

4 years

When students first discover mathematics or logic, they often fall in love with how compact formal notation is. But writing that way makes your writing harder to follow. This advice is great at unlearning bad habits of formal writing.

12

39

162

11

27

196

julesh

@_julesh_

4 years

The unreasonable effectiveness of category theory in getting likes on twitter

12

13

193

julesh

@_julesh_

7 months

In 2007 DARPA launched 23 mathematical grand challenge problems, and number 13 was “Create a [differential] game theory that scales” - on the same footing as “19: settle the Riemann hypothesis” and “1: build a functional model of the brain”

5

37

193

julesh

@_julesh_

5 months

Obviously not everything can be understood with just category theory, although I’m starting to kinda suspect that everything can be understood with a mixture of category theory and statistical physics

14

195

julesh

@_julesh_

5 years

Trying our best

5

30

183

julesh

@_julesh_

3 years

Annual reminder that the first letter of Xmas is a capital χ (ch), not a capital x, same as in (La)TeX and arXiv

7

24

186

julesh

@_julesh_

1 year

Thing I just learned: Monte Carlo methods were invented at the Manhattan Project for studying neutron diffusion in the core of a warhead, and the name Monte Carlo originated as its military code name

Monte Carlo method - Wikipedia

en.wikipedia.org

10

13

185

julesh

@_julesh_

3 years

By Rice's theorem, all nontrivial problems are undecidable, so you might as well give up and have a beer

10

16

182

julesh

@_julesh_

3 years

The question "how can a definition be wrong?" sounds like a real head scratcher, unless you're a practicing mathematician in which case the answer is "extremely easily"

10

16

184

julesh

@_julesh_

6 years

In every n-dimensional manifold M there are a group of people calling themselves "flat M-ers" who insist that they live in an n-dimensional Euclidean space

6

46

180

julesh

@_julesh_

3 years

1

20

181

julesh

@_julesh_

5 years

If I was Stephen Wolfram I'd be extremely proud of my visionary contribution to user interface design (Mathematica predated Jupyter by almost 3 decades), and then just call it a day

7

11

178

julesh

@_julesh_

4 years

Small brain: Mathematicians saying that mathematics is apolitical Galaxy brain: Economists saying that economics is apolitical

8

15

167

julesh

@_julesh_

2 years

The take “ChatGPT is useless because it generates confidently wrong bullshit” is really misguided. There are *many* situations where being able to automatically generate confidently wrong bullshit very efficiently is a massive game changer. Grant writing comes to mind for example

7

15

175

julesh

@_julesh_

3 years

Anyone need a new tensor product?

3

6

169

julesh

@_julesh_

3 years

Periodic reminder that: 1. Haskell is significantly older than Java 2. Haskell is the same age now as Algol was when Haskell was created

5

16

166

julesh

@_julesh_

4 years

Apparently the Facts And Logic crowd think that logic refers to Aristotle-style """"logic"""", which was basically wrong in every conceivable way, and has been obsolete since the 1890s. They can't tell the difference between logic and rhetoric, even though they aren't related

12

10

161

julesh

@_julesh_

4 years

mathematicians 🤝 gamers preorders

1

14

160

julesh

@_julesh_

4 years

Capitalism is a nice idea in principle, but human nature means it will never work in practice

17

163

julesh

@_julesh_

3 years

@kareem_carr Yep, it’s definitely Twitter that needs to change, not academia

2

1

149

julesh

@_julesh_

6 years

You want to stay relevant as a software developer for the next 1000 years? These are 3 major things you should focus on: - Lawvere theories - Kan fillers - Monoidal adjunctions

fogus

@fogus

6 years

45

535

2K

9

33

160

julesh

@_julesh_

3 years

Tired: Giving your 4 year old an iPad Wired: Giving your 4 year old an iPad that's been jailbroken and running a bash shell

5

17

153

julesh

@_julesh_

4 years

@fermatslibrary I think we're told this information in the Silmarillion, it used to be flat until the Remaking of the World when Numenor was destroyed, after that it become curved

1

152

julesh

@_julesh_

4 years

Extremely unpopular opinion: The average quality of pop music has gradually improved over time Obviously the 60s and 70s produced a much higher rate of geniuses and visionaries. But the *average* pop song now is better than the average pop song then

47

15

154

julesh

@_julesh_

5 years

A point from Ostrom's book: A free market is a public good Freeness of a market is an artificial thing that needs to be actively enforced. If you let people do what they like (laissez-faire) they'll tend to do things like forming cartels, that stop the market being free

9

20

155

julesh

@_julesh_

4 years

Reminder that the method of discovering the truth by rational debate is known as "scholasticism". Since it was the dominant method for a millennium before being replaced by science, we have a lot of information about how well it works: it was a wholesale failure. It doesn't work.

11

29

152

julesh

@_julesh_

3 years

At some point all of the talks from Applied Category Theory 2021 were quietly put online:

Applied Category Theory

www.youtube.com

2

42

153

julesh

@_julesh_

2 years

I’m not quite ready to eat my words over repeatedly saying current generation machine learning methods will not lead to AGI; but I’m warming my words up in the microwave

4

12

149

julesh

@_julesh_

3 years

*almost dies laughing*

6

27

148

julesh

@_julesh_

5 years

Integration is hard (from )

3

14

147

julesh

@_julesh_

1 year

Major take of mine: the universe is nice to physicists by being “continuous” in the sense that “small” differences between your model and reality *often* lead to quantitatively controllable errors. In some other fields, like econ, small changes to the model usually blow up

ieva

@HyperboIeva

2 years

I think the joke about how physicists approximate everything is actually a very cool observation about mathematical modelling: In math, you define the model and the structures you're working with: these tend to be precise/exact because you "define" them to begin with... 1/3

6

11

155

18

14

146

julesh

@_julesh_

3 years

Major new preprint time! “Towards foundations of categorical cybernetics” by @mattecapu , @bgavran3 , @AyeGill and me It’s a bit terse because of the page limit, but I think that the thing sketched here is basically the *right* foundation

9

30

147

julesh

@_julesh_

1 year

Who would win? Common sense, or one brackets boi

Jon

@BiggestttHater

1 year

what is y’all most liked tweet?

4K

5K

42K

4

10

149

julesh

@_julesh_

3 years

Common sense says that most applications of machine learning would amount to saving money by replacing human workers with machines. But in practice, it seems that most applications of ML are variants of “bias laundering”: making decisions that would be illegal for a human to make

6

23

148

julesh

@_julesh_

4 years

@PawSob People seeing (f x) : "too many brackets!" People seeing f (x) : "much better"

4

5

146

julesh

@_julesh_

5 years

Forrester, Industrial Dynamics (1961) A very simple model of a length-3 linear supply chain (73 equations), driven by noise from retail sales The system responds to those frequencies in the noise that it resonates to, with a period of ~6 months

2

25

142

julesh

@_julesh_

3 years

My favourite example is the definition of matrix multiplication, I have no idea how you could explain it to someone of the level where I first learned it. It's a major result of 1st year linear algebra to explain where the definition comes from - it's linear function composition

Michael Kinyon

@ProfKinyon

3 years

Do you have a mathematical fact you "learned" in school in the sense that you had to memorize it without really understanding it? This came to mind because many of Math Twitter's grad students are discussing this in regard to their qualifying exams. (1/7)

103

36

315

13

8

143

julesh

@_julesh_

4 years

As it happens the main qualification needed to be a researcher is that you *enjoy* being regularly wrong, confused etc

8

16

141

julesh

@_julesh_

4 years

I had a problem, so I tried using category theory. Now I have a problem in the category of endofunctors

5

10

147

julesh

@_julesh_

7 months

Abstract for an upcoming paper, to be released on the arXiv next week: Reinforcement Learning in Categorical Cybernetics

3

29

144

julesh

@_julesh_

4 years

Yoneda lemma: If it walks like a duck and homs like a duck, then it's naturally isomorphic to a duck

2

15

141

julesh

@_julesh_

5 years

I overwhelmingly think that the human race is wasting its potential. Humans can do incredible things, but the best are being put to work to make money rather than to do anything useful or important

14

16

141

julesh

@_julesh_

1 year

1) This is heavily abusing the technical term “algorithm”, as AI people keep on doing but should know better 2) Using ML for low level code generation is a really good idea, in the spirit of the legendary work in the 90s using genetic algorithms to synthesise “digital” circuits

Google DeepMind

@GoogleDeepMind

1 year

As society becomes more digital, it’s critical to improve the code powering the world's computing. Today in @Nature , we present AlphaDev, an AI system using reinforcement learning to discover enhanced computer science algorithms. How does it work? 🧵

29

443

2K

7

21

145

julesh

@_julesh_

3 years

Over-abstraction is the devil that tempts every mathematician with false promises of power and riches

13

12

138

julesh

@_julesh_

5 years

Assorted links The paper introducing open games: Compositional game theory bibliography: On compositionality: Applying the rising sea: Open games tool demo:

[3S2] Open Games Hack

Statebox Meetup Berlin (part of the 3rd Statebox Research Summit)playlist: https://www.youtube.com/playlist?list=PLE2y3n8EQ6VyuBqCetoDVhRnM2aGGwOozmore info:...

www.youtube.com

3

27

138

julesh

@_julesh_

4 years

And don’t get me started on “computer” “science”, which also doesn’t use scientific method and only occasionally involves a computer

🔥Kareem Carr | Statistician 🔥

@kareem_carr

4 years

Mathematics does NOT use the scientific method. As a discipline it assumes that human reason is the best guide to truth. If experiment and reason conflict, in mathematics we are first and foremost guided by reason. In science, we're guided by experiment first and reason second.

54

66

639

11

16

142

julesh

@_julesh_

4 years

2+2 =/= 4 because 2+2 is a question and 4 is an answer. How stupid would you have to be to think that a question can be equal to an answer?

12

10

135

julesh

@_julesh_

4 years

There is no such thing as a "Harvard study", that's something that only exists in the fantasy world of journalists. Studies are done by scientists, not universities I published a bunch of papers while I was at Oxford, but none of them is an "Oxford study"

3

7

133

julesh

@_julesh_

4 years

Academia is based on proof of work too. We burn a lot of carbon to fly to conferences and tell other people what we’ve been doing

4

16

137

julesh

@_julesh_