Gary Marcus

And I would argue most of the brain is devoted to this kind of adaptive control of behavior. And it would be super cool to build AI like that - as artificial agents...

But is that what AI folks want? Or are they after the ability to perform abstract reasoning and other really intellectual operations better than humans can?

Indeed, what AI folk want is abstract reasoning. They tried jumping straight to it under the Minsky Symbolic AI program, which failed. Maybe we can jump straight to human cognition with ML, but i'm skeptical. I think we need to pass through mouse intelligence. Short jump to human

a. premature argument to say "symbolic AI" failed (using < .01% current compute) when deep learning could have been dismissed in same way in 2009 b. ignores hybrid neurosymbolic models. c. it's not that short a leap, inasmuch as vast majority of mammals did not evolve language

Yes, most animals didnt evolve language. But language evolved in a blink, ~200Kyrs (or 1Myrs, if you think Neanderthal and Denisovan had language). And population sizes were small, and generations long, which suggests it's a very easy evolutionary step for apes to get language.

- it’s not that easy or more primates would be talking; the adaptive advantage is likely huge - but yes the only way to evolve language quick is to already have a genome packed chock full of innate tools that are good for cognition. that was key point of The Birth of the Mind

So, then what are those elements of cognition that brains have that AI doesn't? Is it the architecture for predictive processing? Is it that they incorporate value and meaning through experience?

again my opening bid is the list of 10 things here: https://arxiv.org/abs/1801.05667