big red Chomskian structuralist warning

The principle of generativity states that the goal of a grammar should be to enumerate the span of the structural descriptions of the expressions of a language.

We don’t want to write the same thing many times; generics minimizes code duplication. Therefore, generics!

Let’s implement a simple swap function:

void swap_ptr_values(void *data1ptr, void *data2ptr, size_t datasize) {
}

helper functions

memcpy

Copy datasize bytes worth of memory in the second argument into the first argument. The two arguments CANNOT OVERLAP otherwise, you risk UB.

void *memcpy(void *dest, void *src, size_t nbytes)

memmove

Its memcpy, but it works with overlapping data, and is slower.

void *memove(void *dest, void *src, size_t nbytes)

pointer arithmetic with generics

Unfortunately, given that we don’t know how big a void * pointer is, we can’t do pointer arithmetic against it because it still doesn’t know how big the pointer is. You can’t just add/subtract numbers to char *.

A genetic algorithm is a search heuristic that is inspired by Charles Darwin’s theory of natural evolution.

Its what Grey’s video says. The picking and chucking iterative thing.

Genetic Policy Search involves performing Local Policy Search, but starting from a plurality of initial policies and perturbing the top-k most successful ones (called the “elite samples” \(m_{elite}\)) to generate the next set of starting points.

GenSLMs are a LLM, but genome sequence

1. Take genome sequence
2. Throw transformers at it
3. create “semantic embedding”
4. autoregression happens

This is trained as a foundational model to organize the genomic sequence

Turns out, the embedding space above can be used to discover relations. See proteins can be encoded as hierarchies