There is a use case for every tool, AI being one of them. The first time I really heard about ML in production was DD's Masquerade. No VFX artist I know thinks AI is useless, it's just that image and video generation tools aren't iterative enough for production. Right now, iterating with human direction, and having the AI understand the direction, is impossible.
If we really want to start taking the technology seriously, we should be looking at pipeline integration and modular toolsets, rather than wholesale replacement.
Think of it like like transferring skillsets from claysculpting to Zbrush or Mudbox. Or transferring GoMotion/2d animation skills to Maya.