this may be changing. as we speak. terence tao has recently been working to formalize existing work in lean. and has already found a bug in one of his published proofs. i think there will always ish be a need for mathematical proofs in the contemporary sense, but at this point i don't think its too implausible to suggest a future where the mathematical mainstream moves towards some hybrid mathematical exposition of computationally verified artefact.
https://mathstodon.xyz/@tao/111287749336059662