• Stop being a LURKER - join our dealer community and get involved. Sign up and start a conversation.

Reply to thread

Bill, I appreciate your breakdown of AI's facets. My skepticism about LLMs and math stems from testing their boundaries.


The more challenging the task, the more frequently they hallucinate. I witnessed this while developing an image lighting scoring system - despite implementing explicit anti-hallucination instructions ("No fabricated information," etc.), the system still produced fictional results with absolute confidence.


(yup, I used AI to write a prompt to keep it from inventing answers when it had no answers to give)