I agree with some of your points.
That we reduce a complex system to its components is a tactic that the human mind does to cope with those systems with which we can cope.
If systems are not so composed then we have found no way to cope with them, except on occasion by computer simulation.

A quibble: there are in popular chaotic systems, such as the weather, predictions that you can make.
For the weather they amount to the prediction that climate won't change.
This is not a profound point but climate is something that weather models can predict.

I would say that reductionism fails in practice, (and we do presume to practice!)
(In theory there is no difference between theory and practice. In practice, there is. Chip Morningstar)

In your "New Strategy" slide I think that there is room for the introduction of logic, in the form of predicate logic.

I think that you mean by 'model' something a bit different than what I mean.

There is a confusion that goes back to Newton's time.
The notion that you can compute something in principle was confused with the notion that you could compute it.
The notion or reducibility may be in the same category of confusion.

I am slightly suspicious of AGI even as I am suspicious of "GI".
I think we will have many quirky special intelligences, perhaps to converge sometime on something rather general.