Ask HN: Is fast.ai's "Deep Learning for Coders" still relevant in 2025?
Dear all,
I learned some basic ML from Andrew Ng's Coursera course more than 10 years ago, recently I graduated from the Math Master program and have some free time in my hand, so I am thinking about picking up ML/DL again.
In [Yacine's video](https://www.youtube.com/watch?v=ph6PIchDOcQ), he mentioned [fast.ai's course](https://course.fast.ai/), which I heard of in the past but didn't look into too much. The table of contents of [the book](https://www.amazon.com/Deep-Learning-Coders-fastai-PyTorch/dp/1492045527) looks pretty solid, but it was published in 2020, so I was wondering given the pace of AI development, do you think this book or course series is still a good choice and relevant for today's learners?
*To provide more context about me*: I did math major and CS minor (with Python background) during undergrad but have never taken any ML/DL courses (other than that Coursera one), and I just finished the Master program in math, though I have background and always have interests in graph theory, combinatorics, and theoretical computer science.
I have two books "Hands-on Machine Learning" by Geron and "Hands-on LLMs" by Alammar and Grootendorst, and plan to finish Stanford's CS224N and CS336 and CMU's DL systems when I have enough background knowledges. I am interested in building and improving intelligent systems such as DeepProver and AlphaProof that can be used to improve math proof/research.
Thank you a lot!
What if your goal is to transition into the role of an ”AI Engineer”, would the course be valuable for this?
Yes it's still very good if you want to toy around with it. I did it and in the end I feel like I could do simple models and understand how it works.
If you want a job, it may not be what you need.
If you want to work with ML you likely need the PhD + get lucky.
Remember that the skillset for AI jobs != ML.
If you are interested on it because of AI, it's better to focus on LLMs and more NLP related fields.
Thank you for your suggestions!
Originally I was more interested in ML/DL theory and mech interp, so you can see I was more into theory. Recently, I am also curious and leaning towards learning more about how to build products with foundational models, such as LLM; for instance, recently I got the "LLM's Engineer Handbook [1]".
> If you want to work with ML you likely need the PhD + get lucky.
I do am interested in doing PhD in theoretical computer science, but I'm not too sure about AI PhD as I heard many (not-so-good) things about it.
> If you are interested on it because of AI, it's better to focus on LLMs and more NLP related fields.
Do you have any recomenndations that you have gone through that you think are helpful?
[1]: https://www.oreilly.com/library/view/llm-engineers-handbook/...
I found this to be helpful
https://www.oreilly.com/library/view/ai-engineering/97810981...