Magos Logis Logan @jmw150@qoto.org

Ah... The fraudulent research coming out of MIT and Cornell, and my 4 hours of sleep, are not mixing well.

Neural networks are tensor sensors.

Happy Pi Day!

Hmm.. Who are your corporate overlords, languages?

Rust (4 of big 5 tech)
https://foundation.rust-lang.org/members/

Julia (3 of the big 5 tech)
https://juliacomputing.com/

LLVM (3 of big 5 tech)
https://foundation.llvm.org/docs/sponsors/

Haskell (1 of big 5 tech)
https://haskell.foundation/

Ocaml (financial companies)
https://ocaml-sf.org/

Zig (individuals)
https://github.com/sponsors/ziglang

D (1 financial firm + individuals)
https://dlang.org/foundation/sponsors.html

R (researchers and universities)
https://www.r-project.org/foundation/donors.html

Loops be like, a whole research field.

https://dl.acm.org/doi/abs/10.1145/3453483.3454104

https://dl.acm.org/doi/abs/10.1145/3498682

Hmm... compilers, modules. That is honestly not too bad. Deep learning compiler expertise are in demand because of this kind of stuff.

Seriously just need to call them modules though I think. Tensors do not have non-linear activation functions. In most cases it is just slang for a n-box of numbers.

http://adam.chlipala.net/papers/AtlPOPL22/AtlPOPL22.pdf

Do I care? Eh.. Is the math and programming language stuff nice? Yeah.

I guess this week is compiler week.

A compiler compiler design. It lets you write compiler optimizations as a lot of separate, easy to think about passes. And then the compiler compiler combines them.

https://dl.acm.org/doi/pdf/10.1145/3140587.3062346

I feel like this existed already. It is a little too obvious of an optimization.

I think my top three language picks would be: C++, Python, and probably Scheme. It really comes down to the cathedral and the bazaar.

Even though C++ and Python have commities, Python has an exposed ast and is probably the most popular language. So it is like going to the bazaar. You can find really anything there. C++ has inline assembly and many weird hacks. A large chunk of it is just undefined behavoir. And, a lot of programming paradigms can go and set up shop in C++. Many other languages say they have a unique quirk, only for it to end up in Python and C++ a few years later.

Languages that tell you what to do are cathedrals. Scheme languages have their 9 primitive functions (9 holy primitives) and that is about it to be a scheme. Their committee tries to make RSR7 do everything. But so few of the older schemers actually care.

"Civilization advances by extending the number of important operations which we can perform without thinking about them" - A F Whitehead

Even more automatic memory error detection.

https://dl.acm.org/doi/10.1145/3377811.3380323

Automatically detecting memory related bugs

http://opera.ucsd.edu/paper/Micro04-AccMon.pdf

Neat. Tropical algebra can explain why deep learning systems become more expressive with more layers.

https://arxiv.org/abs/1805.07091

More so differentiating instead of just surviving.

It is kind of like how Tensorflow is faster than PyTorch, but Pytorch is more adaptable at runtime. So it is more of a visible tradeoff.

Julia's system is already going to competed with by Python's Myia. But Julia has a natural advantage of naturally being better designed for speed. So I can see it surviving.

https://github.com/mila-iqia/myia

Rust's Egg, on the other hand, is probably going to be pushed back down by something in C or C++. There is nothing Rust can really do about this.

C++ still keeps a 0 overhead principle, and C is just simpler to work with. These are still languages taught to people in school, as a minor requirement for getting a degree. So there exist competent C/C++ programmers to do the job.

Hmm.. empires have to fall before they are replaced right?

I do not think companies that have "replace Facebook", or something similar, as their mission statement are going to succeed. You have to find what makes your group great.

Julia:
- has reflection and speed
- awesome unique project https://fluxml.ai/Zygote.jl/

Rust:
- idiot proofed manual memory management.
- awesome unique (for now) project https://github.com/egraphs-good/egg

MLIR is going to be so great! Yay compilers! *begins frothing at the mouth*

Me: Please sleep. You are leading discussion tomorrow.

Brain: but I'm booorred.

It has been 2 years of this. The habits have stuck. This is just what I look like at the university now.