• 1 Post
  • 49 Comments
Joined 1 year ago
cake
Cake day: June 7th, 2023

help-circle

  • Re thumb-key do you have recommended tutorials for getting comfortable with it? I found trying to do touch typing tutorials didn’t really help, both because they are generally made for desktop environments and they are geared towards qwerty layout (e.g., get comfortable with home row first etc). I tried forcing myself to use it for a full 24 hours as the concept makes a ton of sense to me, but got very frustrated with myself and then dug into the world of which layout to choose, got overwhelmed, and switched back to whatever this qwerty layout that samsung one ui provides on galaxys.


  • I think that is overly simplistic. Embeddings used for LLMs do definitely include a concept of what things mean and the relationship of things to other things.

    E.g., compare the embeddings of Paris, Athens, and London to other cities and they will have small cosine distance between them. Compare France, Greece, and England and same. Then very interestingly, look at Paris - France, Athens - Greece, London - England and you’ll find the resulting vectors all align (fundamentally the vector operation seems to account for the relationship “is the capital of”). Then go a step further, compare those vector to Paris - US, Athens - US, London - Canada. You’ll see the previous set are not aligned with these nearly as much but these are aligned with each other (relationship being something like “is a smaller city in this countrry, named after a famous city in some other country”)

    The way attention works there is a whole bunch of semantic meaning baked into embeddings, and by comparing embeddings you can get to pragmatic meaning as well.



  • It varies on who does the interview but I push for much simpler than leetcode type stuff- e.g. not puzzle problems but more “design a program that can represent a parking structure and provide a function that could be used for the ticket printer to determine where a new car should park, as well as one that can run upon exit to determine payment”

    Then if they are actually solid we can dive into complexity and optimization and if they can’t write a class or a function at all (and esp if they can’t model a problem in this way) it’s really obvious.