I did not read the paper. Only some comments based on the video: https://www.youtube.com/watch?...
If I understand it correctly; using references to recompute everything in a data dependency manner to save as much space as possible would lead only to space reduction of O(t*log(t)) when compared to space unconstrained algorithm (i.e. an algorithm which uses about as much space as time). This is the old result. They also shift "temporary" variables typically needed during an computation into the computation intermediate results. They combine storing and modification into the same memory using "roots of unity" number systems. I do not know them but a trivial example is swapping two variables using xor without a need for a temporary:
a = a xor b
b = b xor a
a = a xor b
(similar addition/subtraction trick is not used because it needs carry over in general)
Practical usefulness seems rather limited. We often care more about small computation time than small memory usage.
By the end of the year I suspect half of the people reading this won't have a job.
50% unemployment is highly unlikely.
The unemployment rate maxed at 25% during the great depression in 1933.
Don't be scared. You are not affected by CRA if you do not earn money on your open source software.
The CRA regulates commercial activity:
(10) This Regulation applies to economic operators only in relation to products with digital elements made available on the market, hence supplied for distribution or use on the Union market in the course of a commercial activity.
(10c)
(10c) This Regulation does not apply to natural or legal persons who contribute source code to free and open-source products that are not under their responsibility.
Also non-profit organizations hosting open source are not under CRA. On the other side, commercial entities which profit from open source are responsible! These commercial entities are also responsible to publish vulnerabilities they find in open source and also the patches developed. In general, CRA seems to be more good than bad for open source.
More info. for non-layers is here: https://berthub.eu/articles/po...
Cross compiling does not require anything that has anything to do with AI/LLMs anyway.
It does when the target language (like Rust) has ownership and borrowing which strongly favors your program to have a design suitable for linear (affine) type systems. The system must be redesigned to work idiomatically in Rust.
* UNIX is a Trademark of Bell Laboratories.