Auto-generated captions (something I deal with on a regular basis as I do create video courses) tend to be horrendously bad in quality even on major platforms. Especially with tech content (even if you talk crystal clear).
Auto-generated captions that came out bad leaves both the reader wondering what "jean kids" means when you really said "jenkins" and it also de-values the content creator's brand because now you think "how did such crappy captions make their way into this video?".
Now with that said, I'm all for providing human created / vetted captions that are accurate, but that should be up to the content creator to decide.
The platform should never step in and create new content based off your copyright content in a format that you didn't explicitly release.
That's a very good one.
My personal opinion is Youtube should not be able to add captions without your consent because suddenly this changes everything.
What if one day Youtube decides to add a transcript feature where they can turn a video into a decently formatted blog post. Now imagine if you had your own website where you used to make a blog post about your video manually and then uploaded your video to Youtube with a link back to your site.
A move like that could remove a lot of traffic to your site because chances are Youtube would outrank your personal site by a lot in search results and would also give people less of a reason to ever click back to your site.
But this is really similar to audio books vs text. They are both independent mediums and having access to 1 shouldn't guarantee free access to the other. It should be up to the discretion of the content creator (or publisher if they are not self published).