"Will AI become thus capable of original and innovative thought?"

it already has.

alpha go taught itself to play go (a problem space WAY to big to solve with brute force) using no rules, guidance, or human strategy and wiped the floor with top human players by making moves that humans do not even understand.

they just showed it the game, the rules, and the scoring and said "learn to play well".

the rest was done with zero human input on strategy, tactics, framing, theory of game, etc

they thought it was broken at first because it was making moves no one had ever seen before, then it turned out to dominate. it turns out humans never even knew how to play go. our theory of the game has been rendered obsolete. and to this day, no one has any idea what alpha go is maximizing/optimizing/doing.

lots of AI is currently self learning. it's not using human theorems. it's making it's own.