Dan Piponi

Disclaimer:
👽 My opinions are not my own. They're beamed to me by aliens

Current life:
🎮 Epic Games

Previous lives:
🍩 a PhD in Riemann theta functions
💥 many years working in movie visual effects
🎈 some years devising navigation strategies for balloons
🎲 a year drawing random samples from tricky distributions

Likes:
🚴 I like to bike
🏃 I like to run
🎛️ my musical tastes lie towards the electronic end of the spectrum
🚀 I like Andor *and* The Mandalorian

Public Key

npub1x78ms56a50yyrl0t6ztxuksk70fp6n5c3wf9u3cl9k8a7g696t5s9nxn0m

NIP-05 Address

dpiponi_at_mathstodon.xyz@momostr.pink

Profile Code

nprofile1qqsr0rac2dw68jzplh4ap9nwtgt085saf6vghyj7gu0jmr7lydza96gmf9fff

Author Public Key

npub1x78ms56a50yyrl0t6ztxuksk70fp6n5c3wf9u3cl9k8a7g696t5s9nxn0m

Show more details

Last Notes

2024-11-07 16:17:18

https://media.mathstodon.xyz/media_attachments/files/113/442/568/928/579/093/original/e9bdb4ad5aeb7537.png

2024-10-24 16:02:28

Life in the 80s. The ad and the reality.
https://media.mathstodon.xyz/media_attachments/files/113/363/236/889/200/618/original/b59159ab75983900.png
https://media.mathstodon.xyz/media_attachments/files/113/363/237/082/227/938/original/d139bf18a9fd54b9.png

2024-10-22 18:03:51

- reply

I remember those days. Before RENUMBERing tools you'd deal with this issue by doing a GOTO to another block with more spaced out line numbers, and then GOTO back. Or similar. It's like you were forced to write spaghetti code.

2024-10-21 04:26:36

- reply

I was unaware that this kind of "unilateral" text insertion was possible so I don't yet know what to turn off. I'm sure I can if I can find a name for it.

2024-10-21 04:22:32

- reply

Related: I've also noticed that URLs are becoming harder to type in both Safari and Chrome. There's a lot of typing, and more annoyingly, looking at the screen, in order to cancel predictive text.

2024-10-21 04:11:36

I wrote some text for Mastodon. iOS inserted the word "wrong" where I had no intention of writing that word. No warning or anything. It just inserted it. It's repeatable. This is the beginning of the end.

2024-10-20 15:13:25

In Mathematica you can define a function via pattern matching
f[x_]:=x+1
and freely redefine it
f[x_]:=x+2
or write a function as a lambda
f = Function[{x},x+3]
But woe betide you if you redefine it using pattern matching after defining it as a lambda
f[x_]:=x+4
as it'll substitute(!!!)
Function[{x},x+3][x_]:=x+4
In reality what happens is you accidentally do this in code to invert a 1000x1000 matrix and now a simple numerical computation becomes the attempt to invert a symbolic matrix with unevaluated subexpressions and it runs out of memory and you lose all your work...

2024-10-17 16:35:19

- reply

There's nothing wrong with retrodictions

2024-10-16 16:30:50

- reply

Just takes a while to write a reply :)

2024-10-15 15:31:09

I seem to be the only person able to hold in mind simultaneously the two thoughts "LLMs are the greatest innovation in computing in a decade" and "anyone who thinks we can safely use this for the first reply from Bing and Google is an idiot".
It's not hard. You can hold the latter thought without having to think "LLMs are stupid" (which is itself a stupid thought to have) and you can hold the former without wanting to use LLMs everywhere.

2024-10-14 23:11:07

- reply

Yeah, one's so big it can hide fusion reactors a mere 30 years away.

2024-10-14 16:48:45

- reply

Back in the 90s when I was a PhD student at King's College, London, John G Taylor (ie. the erstwhile professor of black holes, quantum gravity, and spoon-bending) created a neural network group in the mathematics department. I'd chat with him and he used to regularly tell me things like "neural networks becoming conscious is just around the corner". I never really took him seriously...

2024-10-13 22:37:04

- reply

It's probably a very common bird. The melodies played by the examples I found on YouTube seem less sad to me than our local bird.

2024-10-13 17:07:20

- reply

Turn up the volume? For me it came out super loud because, I think, SoundCloud normalized the volume of something that was originally fairly quiet.

2024-10-13 16:29:17

- reply

For anyone who's curious here's the bit corresponding to the spectrogram.
(AFAICT there are no built in tools to trim audio on iOS. The voice recorder lives in its own pocket universe. There's iMovie but I couldn't find any way to create a project from a .wav file. In the end I downloaded an app called Audio Trimmer.)
SoundCloud has undergone a lot of enshittification since I last used it but this link should work:
https://on.soundcloud.com/dTq4S3F6aE1H4hX5A

2024-10-13 03:30:04

- reply

I get the same with my iPhone 13 Pro with the latest iOS and Safari. Not just you!

2024-10-13 01:07:44

There is a bird that hangs out near the house that my wife calls the Joy Division bird. It sings the most melancholic descending 3 note melody. The Merlin bird ID app identifies it as a Golden-crowned sparrow and that fits the descriptions I can find online. Sometimes it intersperses a 2 note variation. It almost brings tears to my eyes.
And here's a spectrogram.
https://media.mathstodon.xyz/media_attachments/files/113/297/432/056/365/545/original/f0b66f8370960d46.png

2024-10-09 16:34:17

This, but with chemistry instead of physics.
Has the literature prize been announced yet?
https://media.mathstodon.xyz/media_attachments/files/113/278/426/703/402/671/original/8da0d36b136d139c.png

2024-09-30 17:25:44

In every programming language you find yourself wrestling with aspects of the language solving puzzles that are intellectually stimulating. Wrangling the C++ type system is full of puzzles and I think the sense of satisfaction in solving these is one reason why people still enjoy C++ despite it's many flaws.
Haskell also has many kinds of puzzles like this. But there's one enormous difference between C++ puzzles and Haskell puzzles. The former often seem like arbitrary one off puzzles. When you solve Haskell puzzles it often seems like you haven't solved one single puzzle but something generalizable that occasionally rises to the level of being a "theorem".
So sometimes the C++ puzzles can be fun (in that frustrating way puzzles are), but the Haskell puzzles seem to me of more lasting value. (For one's mind, not necessarily for the code that get's written :)

2024-09-28 22:43:17

- reply

I'll probably watch it when it doesn't cost $25 to stream. I have to admit that I find even the colour palette in the trailer off putting.

2024-09-27 04:26:25

Every so often you read a review that's probably better than the film
https://www.sfgate.com/sf-culture/article/megalopolis-movie-review-19760949.php

2024-09-22 15:14:52

- reply

It's long bothered me that the equipotential surfaces seem conflated with the surfaces of constant |g|. The Google "AI" directly claims they are the same thing, probably just reflecting the fact that everyone claims it. I just figured that if I did the math the two surfaces would almost coincide and that's why people say this - but a priori I can't see why they would be so.

2024-09-21 02:30:46

- reply

Someone found a use for nanotubes? A good use too. Amazing!

2024-09-19 18:01:16

Someone has written a followup to the document on visualizing differential forms I wrote ages ago:
https://web.tecnico.ulisboa.pt/~ist189623/wordpress/2021/09/08/on-the-visualization-of-differential-forms/
Their diagrams are wonderful - much better than mine. They also do some stuff I didn't do, like illustrate how d(a∧b)=±da∧b±a∧db. I never thought of that but it's very neat.
They discuss Lie derivatives which is something I've been meaning to write about myself. In a nutshell, to find out how to diagram a k-form after applying the Lie derivative, you just deform the diagram along the flow of the vector field.
And then there's a section on contraction which I need to think about...
https://media.mathstodon.xyz/media_attachments/files/113/165/508/894/581/366/original/4fb1206978543391.png

2024-09-13 20:29:31

TASCHEN had a half-price sale so I ordered their book of fantasy art. OMG it's the biggest book I've ever owned. Yet cheaper than your average textbook on, say, quantales. No photo can do it justice. But here are a couple of pictures of Giger's concept art for Jodorowsky's Dune.
https://media.mathstodon.xyz/media_attachments/files/113/132/119/958/424/568/original/1364361bc63ea709.png
https://media.mathstodon.xyz/media_attachments/files/113/132/120/980/153/266/original/03be824941d979d2.png

2024-09-05 03:55:57

- reply

oh! I regularly search arxiv.org for SETI and that has SETI in its list of keywords but it didn't show up.

2024-09-04 18:16:45

- reply

TBH we've had 55 years since then to see that nuclear power has a good safety record but I don't think it was obviously safe back then.

2024-08-26 23:15:11

- reply

You can go a step further, literally. Convert from inches to cm by taking two steps. So 21", say, is around 55cm (53.34cm to be precise).

2024-08-14 02:56:06

- reply

Spent forever trying to get a camera to charge at the weekend before realizing it needs 9V USB and I'd thrown out that charger because I thought they were all the same.

2024-08-01 02:44:44

This is a really nice expository paper by Henry Cohn "From sphere packing to Fourier interpolation":
https://www.arxiv.org/abs/2407.14999
It gives a surprisingly readable account of work by Fields Medalist Maryna Viazovska.
I want to mention one particular theorem that's easy to state and has nice connections to signal processing.
Suppose you have some function f (with some technical side conditions) with Fourier transform F. If we know that if F is zero outside of a certain interval centred on zero then Nyquist-Shannon tells us we can reconstruct f perfectly knowing only ...f(-2), f(-1), 0, f(1), f(2), ... . All of our audio processing, and much of our image processing, is built on this assumption. In addition, given any choice of values of f(i) we can reconstruct a corresponding f.
But here's a weird and wonderful variant: Suppose f is even (satisfying certain technical conditions about decaying fast enough at infinity). Now also suppose we know f(0), f(1), f(√2), f(√3), ... and also we know F(0), F(1), F(√2), F(√3), ..., then in this case we can still perfectly reproduce f. We can also reconstruct a suitable f given any choice of f(√i) and F(√i) except for one linear constraint coming from the Poisson summation formula: we need f(0)+2f(1)+2f(2)+... = F(0)+2F(1)+2F(2)+... . This looks like something that should have been known classically but it's a 21st century result due to Radchenko and Viazovska. The proof makes use of my favourite functions, the theta functions.
There's an amazing connection between Fourier transforms and sphere packing and an n-dimensional variant of this theorem allows us to prove that E₈ and the Leech lattice are optimal lattice packings of spheres in 8 and 24 dimensions.
https://media.mathstodon.xyz/media_attachments/files/112/884/419/866/427/114/original/b160126d48a5cd5f.png