if by "conceptual developments" you don't mean "a total paradigm shift to some non-probabilistic architecture that hasn't been discovered yet", then yes, they a

if by "conceptual developments" you don't mean "a total paradigm shift to some non-probabilistic architecture that hasn't been discovered yet", then yes, they are a hard constraint of current LLM architectures