[tomrod]

I'm glad an AI researcher is posting their personal thoughts. As a fellow researcher and developer in the AI space, I conceptualize AGI as a self-domain-learning entity that is able to manage itself and understand how to affect/intervene its actions and effectual consequences (humans don't reach this on a general level until early teens to mid twenties, if ever). Walking is a solid example of this at a specific level. Singleton objective functions and patterns across probabilities of states are insufficient -- there needs to be ongoing interpretation and architecture extrapolation.

In short -- current AI approaches are great at interpolation and still perform poorly at most extrapolation exercises that a human can take on.