It's certainly possible (e.g. take a hash of the first few frames of the ad and you can detect it pretty much anywhere and cut the right amount out of the stream).
But it's a lot more involved than just hiding an element on a webpage or blocking the same bit of JS every time.
And while I can see ways to automate it (take two streams for different users, compare differences, etc), it will likely end up being quite intensive on resources.
The only long term solution is "stop using Youtube". We need some fediverse style P2P replacement, where we pay for the videos with our outgoing bandwidth, and we're not there yet. Being a trillion dollar corporation sure does give you a lot more options in how you host things.