It took nature four billion years to invent humans, who are actually - if we're honest - pretty terrible as an example of workable AGI.

In fact what nature invented was a persistent colony organism with external memory.

Wild solo humans are only a little smarter than wolves individually, but being able to share and externalise invention and learning created a massive advantage.

Humans are successful because although only a tiny minority of individuals are any good at invention, the fact that information persists and is shared means the entire population benefits.

The problem for AI is modelling the learning and invention process. Classifiers and recognisers are getting better, but they're not really learning in the human sense, which is a combination of abstraction, mimicry, and occasional invention.

IMO there's no chance of AGI developing until there's a persistent, transferable, abstracted model generated as an output from classifier systems and other learning machines which is a symbolic - not just a statistical - summary of the learning.