I tested something simple with an LLM today.
Give it the same concept with different instructions, and the answer can change dramatically.
Then give it information it didn't previously have, and it can use that context almost immediately.
It made me wonder: How much of what an AI "knows" actually comes from training, and how much comes from the context we give it?
@IGN It's interesting how adaptations balance the book's detached historical style with dramatic arcs for TV. Which change in the finale felt most necessary versus the one that shifted a character's arc the most?