I am extremely disappointed in this choice by Instructure.
I use captions to follow video content, and I want to re-iterate what others have said: auto-captions are not a high-quality way to make videos accessible, in much the same way that machine translation is not a high-quality way to provide material in other languages.
Yes, it is better than doing nothing. Yes, I use the auto-captions when they are what is available.
"Do nothing" is not an appropriate default stance from which to begin deciding which accommodations are necessary in this day and age, particularly in an education-related space and from an institution rather than an individual doing something "on their own time". It's particularly disconcerting if, as others have said, Instructure took the time to do this properly in past years and is no longer willing to accommodate its non-audio-consuming users in this (pretty straightforward) fashion. This makes me worry about their continuing commitment, as a company that provides an LMS used my many institutions committed to serving people with disabilities, to prove accessible tools and services to all users of their product.