This argument sounds very much like the arguments I remember hearing in the early 2000s about open source. "Sure, Microsoft is bad, but can random hobbyists on the internet produce something of the same quality?" "There's just no way that this Finnish dude wrote an entire operating system from scratch, he must have stolen work from this commercial UNIX that took years of paid engineering work to develop." "MySQL is good for people setting up websites at home, but professional users are going to prefer Oracle."
Much like with open-source software wasn't so much about it being free as changing the development model (you could see the source, send in patches, and have a direct discussion with the engineers who wrote that code), one of the more interesting changes is how much research progress now happens via preprints or even just blog posts. There is definitely a valuable role for journals as the gatekeepers of what research you should pay attention to and the facilitators of communication, but like account reps that insulate engineering teams from customers and help prioritize what's important, in practice they aren't working well enough any more to justify having them.
(And the implication, then, is that this work is hardly menial; you want a person actually familiar with your discipline to triage bugs or sort through preprints. And it turns out you're in a better spot to find or train up those people than the middlemen are.)