https://researchgate.net/figure/...

I think this model discounts the idea of "representations regressing to the world as ground truth" or the "world as its own best model" in research related to embodied intelligence. I just think researchers tend to overlook state-free approaches because everything is DNN now.

I can't pretend to understand everything they're talking about, and AI winter may be coming eventually, but its still well worth the read simply because of the approach they take merging neurobiology with ideas from computational intelligence.