I think multidisciplinary knowledge is unnecessary, especially now with deep learning where there is no feature vector design. Even the DeepMind discoveries in

I think multidisciplinary knowledge is unnecessary, especially now with deep learning where there is no feature vector design. Even the DeepMind discoveries in various fields always feature the same people that I doubt know deeply about protein folding or similar stuff. Even when you look at computer vision research and how all the sophisticated methods became unnecessary when NNs came to dominate shows the same thing. I remember having to learn about dependency parsing, part-of-speech tagging, named entity recognition, entity relationship inference, document summarization and a bunch of sophisticated modelling. Combining all of that to get to high level tasks like machine translation or question answering or even summarization (some methods pruned the dependency tree to get a summarized sentence) was difficult. Look at transformers disrupting the NLP. There is no concept of dependency tree, no need to do POS tagging, it's not even necessary to think about that when making a machine translation system. People were figuring out how to build better and faster dependency parsers, POS taggers etc. Domain knowledge was massive and it became redundant with the advent of transformers.