The problem I think is that current LLMs maybe not complex enough to accept all stimulation.

Current LLM systems are more like simulation of the stimulation, a conclusion rather than a exploration.