At the risk of others trying to kill me, I’m going to say what my hope is.
When ML stated, the cameras then, photo or video, clearly were enhanced by the ML technology.
Today, many cameras offer us what we need, eg focus peaking, focus bracketing, exposure bracketing, exposure over/under warning etc etc.
What remains missing, IMHO, is a scripting engine built into the camera, so the user can ‘customise’ things via scripting, eg achieving more automated/complex workflows.
Bottom line: if there is to be a ML 2.0, eg for Post EOSM mirrorless cams, I would put scripting at the top of the list of things to achieve, ie by the code gurus that understand interfacing/linking with the Canon side.
As I say, just my humble thoughts.