this post was submitted on 01 Sep 2026
5 points (77.8% liked)

AskBeehaw

3034 readers
2 users here now

An open-ended community for asking and answering various questions! Permissive of asks, AMAs, and OOTLs (out-of-the-loop) alike.

In the absence of flairs, questions requesting more thought-out answers can be marked by putting [SERIOUS] in the title.


Subcommunity of Chat


This community's icon was made by Aaron Schneider, under the CC-BY-NC-SA 4.0 license.

founded 3 years ago
MODERATORS
 

I know very little so please don't mind a lot of stupid questions. bee emoji

I want to do my masters in AI/ML but afraid that it will be difficult for me to understand. I am not a bright student and struggle with understanding maths a lot. how much maths does an average AI/ML person do in their job? what things I should keep in mind while learning AI/ML?

does whole AI/ML landscape has been taken over by LLMs? I have seen a lot of people complaining about CEOs shoving LLM every where instead of just a simple ML model.

what does AI/ML people do exactly? I mean what are their roles? does the most important task for an AI/ML person is to decide which model to use or is it preparing the data or understanding the requirements?

you are viewing a single comment's thread
view the rest of the comments
[–] swicano@programming.dev 4 points 3 days ago* (last edited 3 days ago) (1 children)

I work in AI/ML, and it's such a hilariously broad and meaningless field I don't even do remotely the same thing as my coworker in the same team hired on the same day. They work in agentic llm swarm research I work in CV/RL. Here's three papers I had to read recently, and not fully understand but extract usable insight from (why did they do what they did and can I use that for myself)

Upsample anything

Dinov3

~~LTGNN line transformer graph neural networks for robust line feature matching~~ just kidding this one isn't on arxiv

It takes me a couple hours to deeply read a paper, I'm not like, getting 100% from it till like the 4th read through and a couple pages of notes, or reading a couple background papers (such as the JBU paper for UpAnything). Try reading one or two and if it seems interesting then maybe that's an area in AI ML that might be interesting to you.

Remember the people doing the most math are the people doing ai ML academically in a masters or PhD, us commercial users tend to do less math and more "this new model came out can we apply it to our problem"

[–] heisenberg@beehaw.org 1 points 2 days ago (1 children)

I think I have to get used to reading scientific papers. could you tell me where to start, I mean some papers which doesn't require a lot of pre-knowledge?

[–] swicano@programming.dev 3 points 2 days ago

Unfortunately, the field is pretty brutal about just assuming you are fully caught up to everything in the field, or that you will read the references for a given statement. You often don't need to know all the math, but at least have a general idea of what some concepts are. I've been part of 'journal clubs' for the past 5 years where the only really goal is to stay abreast of the changes going on so that a new paper doesn't send me down a rabbit hole of reading a referenced paper, which also has unknown terms, and read those references, etcetcetera.

I'm not in a great position to give useful advice due to that. it might be a good approach to find a youtuber who gives a 20-30 minute review/summary of a single paper that looks interesting, and of the video makes sense, then try to read the paper, and if you get stuck, see if the YouTuber has a video on the concept that got you stuck.