Which means that copying birds is fine, you just have to know what parts to copy (mechanisms to control direction in all 3 axes) and which to ignore (flapping wings isn't practical given the power sources and construction materials we have, but soaring on stationary wings is a good place to start).
Which in turn means that saying "Just copy nature, they've been doing this for 1bn years" or "don't bother copying nature, we need different stuff" both aren't very useful.
The analogy between acheiving AGI and achieving flight is invalid on its face and yields no useful information. Nothing can be drawn from the latter about the former.
Flight is mechanical. AGI is not. However, until now, ML and other AI techniques have focused largely on mechanics (stuffing data into an algorithm), which is why such analogies with nature are tempting. But, that approach won't suffice for AGI.
The breakthrough for AGI will come through an approach that backs away from brute force and attacks the problem from a meta-perspective; building on the core blocks of what intelligence is. The approach will be relatively simple, once the core foundation of the integrative learning process is understood, which is the much more difficult task. Execution will then involve establishing a base case that implements, not learning, but the capacity to learn. This will then be replicated across generations, compounding exponentially until what we recognize as AGI emerges.
As well, the required computing power is likely to be significantly less than is commonly believed. This raw brute-force ML-style approach again misleads us here.
I do agree that different approaches are needed.
In the case of flight it was the realization that the airfoil is important and the rest is window dressing.
https://en.m.wikipedia.org/wiki/Fluid_dynamics
Nothing to do with origins of flight but interesting and related to why flight works.
In fact, cognitive science largely split off as a separate field but there's a school of thought that algorithms and big data are only going to take things to a certain point and you're not going to get to things like fully autonomous vehicles under general conditions without different approaches that involve better understanding the physical world.