[kragen]

We haven't found anything in nature that resembles reverse-mode automatic differentiation, either in evolution or in neuroscience.

[mountainreason]

What would that look like. Is there an implementation of the finite difference method in biology somehow?

[kragen]

I'd be surprised.

[mountainreason]

Yeah thanks, me too. So I'm wondering what point 22 is getting at. How could that even be possible?

[kragen]

Biology can use discrete methods; we have discrete neurons (and other cells), discrete chemical species like ATP, discrete amino acids, discrete genes made of discrete nucleotides, etc.

[mountainreason]

Fur sure. But I don't get how that leads to an expectation that our biology is computing and utilizing derivatives in a computational graph.

Point 22 seems to imply that the other finds it notable that that isn't happening.

[mountainreason]

Author*