Language models can solve any riddle as long as they're taught the solution or given the tools to solve it. A human child cannot solve a Riddle either if they are not taught enough language. A human child raised with no humans is basically a wolf. A child raised to speak Russian cannot solve an English riddle. We as humans are insanely constrained by language barriers, beliefs and our meaty minds, LLMs are not. LLMs in their current version aren't an AGI but they can grow to get there in time as long as we keep improving them.