its 2025, no one use RNN and LSTM. For computer vision, CNN and U-net are also outdated.
What do they use then?
1 Like
RNN and CNN are generic categories. RNN is Recurrent Neural Network, which encompasses any kind of sequence model. So that would include Attention Models, which are the basis of all LLMs. So I think saying “no one uses RNNs anymore” is an inaccurate statement. Maybe analogous to the famous Yogi Berra quote: “Nobody goes there anymore. It’s too crowded!” 
CNN means Convolutional Neural Networks. I don’t know what else people may use for computer vision these days, but I’ll bet that the underlying technology is based on convolutions. E.g. YOLO is a CNN.
The larger point here is that from a pedagogical standpoint, it’s never a waste of time to understand how a field evolved. Math is the perfect example: we start by learning things that the Greeks figured out thousands of years ago. Sure there’s more recent stuff to learn, but the work of Euclid and Pythagoras is still relevant. It may be 2000 years old, but it still matters.
3 Likes
If you have suggestions for different topics in Sequence Models, please state them.