Predictive Coding Approximates Backprop Along Arbitrary Computation Graphs (2020)

Also, most recent state of progress: Predictive Coding Can Do Exact Backpropagation on Any Neural Network (2021)

https://arxiv.org/abs/2103.04689

here's a link with reviewer comments https://openreview.net/forum?id=PdauS7wZBfC (praise to openreview!)

AstralCodexTen (formerly SlateStarCodex) has discussed this here - https://astralcodexten.substack.com/p/link-unifying-predicti...

He mostly points to this post in LessWrong - https://www.lesswrong.com/posts/JZZENevaLzLLeC3zn/predictive...

If backprop is not needed, would this finding make automatic-differentiation functionality obsolete in DL frameworks, allowing these frameworks to become much simpler? Or is there still some constant factor that makes backprop favorable?

Needs [2020] in the title. Interesting work nevertheless.

This (and its follow up papers) have already been discussed multiple times here. Don't have the links handy though..