But as you point out - computer scientists don't know what people want _either_, so.. predictive power itself is not necessary for a thing to work.
But _working_ isn't sufficient either. Facebook _works_. As I understand it is WILDLY profitable. But the way it works is a matter of economics and social sciences. Building a social networking website is a SOLVED PROBLEM in computer science. (yes, I realize they're doing some creative research on scaling it up).
Someone could clone Facebook tomorrow, and offer it for free, right? But.. could they charge $10/mo for it, and say "Facebook but without the spying and advertising?" - Probably not.
I'm not saying computer science is useless here. New algorithms WILL be needed. Bitcoin or any form of distributed ledger seems extremely unlikely to scale as needed, for instance. Perhaps provably unable to scale as needed - I'd welcome a paper proving this with solid math / computer science (and there probably is one, I don't keep up).
But what IS necessary is a working economic model that is not the current economic model. You're saying that nobody can predict the future (true), so just try a bunch of stuff - so far, so good. But then when you find something that works.. how do you pay for it? Nobody in computer science is trying to figure it out... that is a consideration "for later" - the point the article is making! What happens "later" when money needs to be made.. well.. advertising and surveillance. That is literally the only answer anyone has come up with.
That is why, FIRST, predictive or not, someone (probably in the social sciences) needs to come up with the idea(s) about how it _might_ be paid for, that don't include advertising.
The article was a response to the big decentralized web conference "web 3.0" that happened recently. It was all CS people, technologists who got together to enjoy talking about technology. But economics is what will shift this, and _innovation_ perhaps will be needed in economics. And other social sciences - nudging, gamification, etc.