Someone who has always loved looking at the space between different AI techniques, this time I went a little deeper into the memory trade-offs between RNNs, Transformers and SSMs. I found it interesting because once you start looking at these architectures through the lens of working memory, a lot o
