I would say the problem is not that academic articles are worthless, but rather that their primary worth is only in the aggregate.
The $5000 piano is a misleading example here. You couldn’t possibly sell academic papers one by one. You can’t have each academic paper cost $35 to access individually.
Any pay-per-paper setup would encourage academics to read as few papers as possible. Even just skimming to see what papers are remotely relevant would cost you hundreds of dollars, and might give you nothing useful at all. You might need to skim a hundred papers to find the five that your actually want; there’s no way anybody can pay $3,500 for five papers.
So the papers are only valuable when you can access hundreds, thousands of them at will. That gives you clear value.
But what that means is that the player with all the power is the publisher, the aggregator — not the author.
One author writes one paper; one paper is indeed practically valueless.
The publisher is in a very powerful position, where he can take a bunch of work that almost nobody would pay for, and turns it into a resource that pretty much everybody will pay for.
That’s the imbalance that needs to be addressed.
Perhaps the answer is for the academics to structure themselves into an aggregate, rather than an ocean of independent individuals.