Originally Posted by
indufan
Why would AI make up any rules? Humans mostly due it due to a lack of knowledge and understanding. I am under no pretense that AI would be perfect but humans fall far below the perfect standard.
Yes, still an open question whether humans or a chatbot would be better at following policy, but the Air Canada case was exactly about a chatbot making up rules (in particular it said a ticket would be eligible for a refund when it wasn't):
https://www.theguardian.com/world/20...hatbot-lawsuit
Also, many large language models (LLMs) hallucinate facts on a regular basis and are also susceptible to adversarial prompts (
https://llm-attacks.org/) which could also be problematic for a chatbot that's allowed to do things like book you tickets or upgrades.