As others have said, the fundamental challenge is that teaching, as usually defined, doesn't scale like that and that turning it into a non-interactive process removes much of what makes it more effective than just giving out a textbook.
Part of the problem is that many people take undergraduate lectures as a reference point and then think "well, there's no reason this couldn't be recorded, have interactive quizzes added, and then be better than physically attending a lecture of 200 people". That isn't wrong per se but mega lectures are barely teaching as it is and we should not be seeking to replicate them!
I know people bang on about the (much misunderstood) 2 sigma problem but I'm going to do it as well. Regardless of the exact result, we know a few things:
First, we know that intensive 1:1 tutoring moves the mean performance up by two standard deviations.
Second, we know from other research that there is a correlation between class size and performance.
That is obviously linked in that as classes get smaller, teaching styles can get progressively closer to 1:1 tuition. A class with ten people in it has room for a substantial amount of brief 1:1 interventions, especially if there are no behavioural problems in the group (so the teacher is not diverting massive amounts of time to classroom management) whereas a class of 30 makes that much more difficult.
Conclusion from that, plus the fact that undergraduate classes are often taught by research experts with no teaching nor mentoring in pedagogic technique, is that the undergrad chalk-on-board lecture with 200 students is at the far end of effective use of time from 1:1 tuition. So if tuition is the most "taught" of the learning styles, a large lecture is all the way over on the opposite end with "just read a textbook".
In fact, universities know this which is why they supplement lectures with grad-student led problem solving / discussion sessions, seminar groups, and small group tutorials. Some universities, for some subjects, use only small group tutorials (e.g. Oxford colleges for many humanities subjects) and lectures if they exist are very much optional. Most universities do not have the resources to do very much of this except for more advanced subjects but this has nothing to do with these techniques only being appropriate for such subjects - small group classes are also the most effective way of teaching Intro Calculus or Introduction to the Novel. Arguably they might be even more important in that context.
So the modern formulation of the two sigma problem to me is as follows: given that we can currently duplicate the (very bad) mega lecture format using technology, how much of the performance from actual teaching (the purest form of which is 1:1 tuition) can be captured using something less resource intensive?