Ask HN: Year Deep Dive in ML

1 points by TheAlchemist 19 hours ago

I'm planning to spend significant time this year on ML and I'm looking for advice, feedback from people that did it successfully.

I have quite a lot (>10Y) of experience in software and everything data already, and at some distant point in the past I did play with NNs a bit - I have a decent (for a non practitioner) understanding - but that was before LLM, transformers.

I'm not necessarily looking to learn things that could get me a job in the field in the short term, but much more on developing a deep understanding of underlying concepts, play with new architectures.

I watched some Yann LeCun talks recently and I quite agree with his view that current trajectory of improving LLMs is a waste of time. It also makes me think about the Capsule Networks (which were looking promising last time I checked, but apparently disappeared completely by now).

So: 1. What resources would you recommend 2. Any advice from your 'journey' 3. What are good forums where one can ask questions