Hi HN,<p>This is a question I've found myself wondering a lot recently. There's obviously been lots of recent progress in the state-of-the-art for LLMs and generative AI, but is this progress trickling over to other areas of AI such as machine/deep learning?<p>If so, what are some resources to get up to speed quickly?
It’s good and useful. Some use cases in NLP have moved over to using LLMs, many use cases in NLP/vision are basically using pre trained models and adding some simple similarity search or simple classification on top of it. Then you have a bunch of deep learning and non-dl methods for forecasting, tabular data, search ranking etc.
Generative AI is is a subset of deep learning which is a subset of machine learning.<p>Note there are a huge number of other ML approaches, two that are useful in tabular data (where LLMs tend to fail) are<p><a href="https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression" rel="nofollow">https://scikit-learn.org/stable/modules/linear_model.html#lo...</a><p>and<p><a href="https://scikit-learn.org/stable/modules/ensemble.html" rel="nofollow">https://scikit-learn.org/stable/modules/ensemble.html</a><p>of which XGBoost is still winning competitions.<p>There is also the “old AI” based on logic that is doing well, see Donald Knuth’s notes on developments in SAT<p><a href="https://www.inf.ufrgs.br/~MRPRITT/lib/exe/fetch.php?media=inf5504:7.2.2.2-satisfiability.pdf" rel="nofollow">https://www.inf.ufrgs.br/~MRPRITT/lib/exe/fetch.php?media=in...</a>