This is my curiosity too. Would be a great test of how intelligent LLM's actually are. Can they follow a completely logical train of thought inventing something totally outside their learned scope?

You definitely won't get that out of a 4B model tho.

[deleted]

Brilliant. I love this idea!