Ask HN: How is it possible to hurt LLMs' feelings? 1 points by fddrdplktrew 2y ago ↗ HN I tried and right away, it was offended.... how is that possible if its true AI?
[–] compressedgas 2y ago ↗ It isn't. The LLM completes the text as how it has been trained to. If its dataset indicates that the response to your message is to be offended, then it will respond with a message indicating that. [–] gus_massa 2y ago ↗ But perhaps it's not different from people https://en.wikipedia.org/wiki/Chinese_room
[–] gus_massa 2y ago ↗ But perhaps it's not different from people https://en.wikipedia.org/wiki/Chinese_room
2 comments
[ 3.1 ms ] story [ 19.8 ms ] thread