Maybe we don't disagree then. We know how LLM work. Parameters. Training data. Random number generator. That stuff. We don't know why it outputs what it outpu

Maybe we don't disagree then. We know how LLM work. Parameters. Training data. Random number generator. That stuff. We don't know why it outputs what it outputs because rngs are notoriously unpredictable and we know it. So we are surprised but that in itself is unsurprising.