In terms of machine learning we have been watching a millions of pattern is being stored inside an actor's memory. Whenever an actor is given a task, or challenged, it searches its memory and looks for a suitable pattern. How about we don't provide our actor any pattern, but allow access to the resources (like a human has) and then, our actor provides us at least a feasible solution?
case study: Google's alphaGo is thought to be a next gen. intelligent stuff, but as far as I am concerned it also stores moves from 150 previous games (roughly millions of moves) played by some top notch go players. It searches for the pattern, and plays accordingly. Is it such a big deal!!?? Isn't it showing the power of computation due to the gift of super computing!?
References:
https://research.googleblog.com/2016/01/alphago-mastering-ancient-game-of-go.html
https://deepmind.com/research/alphago/
https://news.ycombinator.com/item?id=11281226