Transformer gives you function \(F: [X] \to [Y]\). Anything that can be totally ordered can be learned. So it’s pretty general.

I just made the slug for these URLs random since I’m sick of caring about them. Search is good enough to not have to care.

Unknown among my community, Fabrice Bellard is someone I aspire to be a third as productive as.

I saw some people on Twitter recently talking about work that’s similar to stuff I’d done last year (~April 2019). For a moment, I felt annoyance that I wasn’t cited, but then that quote about standing on each other’s shoulders and not each others’ toes came to mind, so I decided to blog about the paper since my marketing clearly wasn’t good enough.

After years of putting it off, I finally prepaid a tutor to force me to actually read about differential forms.

The one PyTorch framework that doesn’t drown me in bureaucracy, though it’s buggy as hell since features seem to be added with about 10 seconds of consideration.

Tries to give the ultimate top down view of physics, as of 1960. Thankfully it’s not The Road to Reality.

• He’s honest about intentions (this book is for aspiring physicists)
  • Sucks if that’s not you I guess
• I feel sorry for the students who took it for a grade
• It blows my mind just how well certain things are known (like the fact that the angle between hydrogen atoms in water is 105 degrees and 3 minutes)

• Nesting function calls is ugly IMO compared to dot chaining.
• I like that broadcasting is explicit, I think Mat + Vec should be a type error.
• I wish it had auto-currying. I wish all languages did.
• I wish there were tagged unions.
• Multiple dispatch is slick, and a natural fit for math, though I wonder how much/if specialization issues like what Rust’s been plagued with are issues in Julia.
• Custom operators are nice
• I don’t like begin/end, I’m fine with curly braces or something, though they do look pretty on generics (but so do square brackets)
• Julia used * for string concatenation because strings are a monoid, so they seem to care about mathematical convention. Why not use arr(i) to emphasize that array access is a function from the naturals to your input set?
• Don’t care about 0 or 1-indexing.
• Zygote/Flux is so cool.
• Chris Rackauckas is a mensch.
  • His arguments about how Julia enables building good packages, which is what scientists want even harder than most developers, will be a big advantage.
• I don’t like ternary operator. Make conditionals an expressions.
• I think I’d still like it if Swift with some sort of value-based generics was the default ML language (or really a go-to language in general). If only it wasn’t so tied to macOS.

My friend, Adrian Albert, was killed in a car crash 2 months ago.

I was sitting there, as I said, and had been for several watches, when I came to me that I was reading no longer. For some time I was hard put to say what I had been doing. When I tried, I could only think of certain odors and textures and colors that seemed to have no connection with anything discussed in the volume I held. At last I realized that instead of reading it, I had been observing it as a physical object. The red I recalled came from the ribbon sewn to the headband so that I might mark my place. The texture that tickled my fingers still was that of the paper in which the book was printed. The smell in my nostrils was old leather, still wearing the traces of birch oil. It was only then, when I saw the books themselves, when I began to understand their care.

His grip on my shoulder tightened. “We have books here bound in the hides of echidnes, krakens, and beasts so long extinct that those whose studies they are, are for the most part of the opinion that no trace of them survives unfossilized. We have books bound wholly in metals of unknown alloy, and books whose bindings are covered with the thickest gems. We have books cased in perfumed woods shipped across the inconceivable gulf between creations—books doubly precious because no one on Urth can read them.”

“We have books whose papers are matted of plants from which spring curious alkaloids, so that the reader, in turning their pages, is taken unaware by bizarre fantasies and chimeric dreams. Books whose pages are not paper at all, but delicate wafers of white jade, ivory, and shell; books too who leaves are the desiccated leaves of unknown plants. Books we have also that are not books at all to the eye: scrolls and tablets and recordings on a hundred different substances. There is a cube of crystal here—though I can no longer tell you where—no larger than the ball of your thumb that contains more books than the library itself does. Though a harlot might dangle it from one ear for an ornament, there are not volumes enough in the world to counterweight the other.”

I got this from the headings in the Wikipedia article, so this is a conservative lower bound.

I had a hard time remembering this¹, almost as bad as Type 1 vs Type 2 (which are still the worst terminology I’ve ever seen).

Lots of languages have packages for benchmarking. But writing code to do that is a pain when you just care about getting a sense of how fast it is as a whole.

I found out about this guy yesterday. He was a famous Canadian folk singer¹ who died in the accident that led to planes having a special section for people who’re supposed to help out in emergencies (with extra legroom as a nice perkquisite).

https://www.outsideonline.com/2395803/snakebite-antivenom-tim-friede