ML Algorithms vs LLMs

AI Discussions
1

Hello everyone!

There’s one thought that has been stuck in my mind for a while and got intensified now that I came back to ML studies.

Do you think traditional ML algorithms got less relevant with the quick advancements of LLMs?
At Supervised Machine Learning: Regression and Classification, there is one example of how Google News make use of a unsupervised algorithm (clustering) to categorize news that are similiar between them.

That makes me think though, if nowadays it wouldn’t be more straghtforward to just leverage an LLM to do that job.

I know there might be cost effectiveness concerns that justify the use of a specialized/simpler algorithm such as clustering, but I’d definitely like to hear more from people who are more experienced.

Thank you!

2

If you want to know how LLM’s work, I recommend you start with linear regression.

1 Like
3

I don’t think there is really a versus, but rather that they are complementary sciences. It’s mainly about thinking in terms of deterministic systems versus probabilistic systems.