Ask HN: Should I (re)learn math to transition to ML? | Hacker News Reader

Ask HN: Should I (re)learn math to transition to ML?

2 pointsbtop4 comments

TL;DR: How to learn math for AI on the side?

Hey, I am currently a senior software engineer working mainly with backend systems. As I have to work another 30 years or so, I cannot deny that ML knowledge is more and more important nowadays and will become even more in the future.

I do not think that just learning the current frameworks will do me any good in the long run. Understanding the math and tech under the frameworks seems important to me for future growth.

However, my last math class was yeeears ago. Now my time is limited.

Is there a suggested way to learn the basic math to understand current(!) ML trends?

View on HN

No, you don't need to transition to ML, you don't need to work on ML, you don't need to learn ML (unless you want to of course). There's no ticking clock, you don't need any of this in order to have a good career.

> As I have to work another 30 years or so, I cannot deny that ML knowledge is more and more important nowadays and will become even more in the future.

I think you're confusing being a user of ML tools with being someone who works on ML tools/systems/products. The landscape of available tools has really changed in the past year; the landscape for working in the field hasn't really - there aren't significantly more jobs, it's not much better or worse, it's not a job that everyone will need to do some day, it's just an okay sub-field of software engineering and/or research.

> I do not think that just learning the current frameworks will do me any good in the long run

Why? This feels like saying you want to learn assembly because you have no faith in C. Abstracted tools will be all there is someday soon, the amount of people working on the math side of things will probably be highly specialized. You could learn this stuff now, but you have to consider who you're up against and what you'd actually do with that knowledge. If you write your own LLM and custom accelerator, you're still building the same product as the guy who just calls the OpenAI API.

I agree, although there is always some minority that will know things at that level. If you think your job (or if your current job) will involve optimizing ML libraries or a large ML project that's over budget, or under performing, having an idea what's being performed under the API can give you a quick insight into what can be changed.

Check out Coursera, there's an ML intro course by Stanford that will walk you through all the relevant math https://www.coursera.org/specializations/machine-learning-in...

Reply on news.ycombinator.com