4 ms·
Ask HN: How is it possible to hurt LLMs' feelings?
I tried and right away, it was offended.... how is that possible if its true AI?
- compressedgas 3y agoIt 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 3y agoBut perhaps it's not different from people https://en.wikipedia.org/wiki/Chinese_room https://en.wikipedia.org/wiki/Chinese_room