Leo Chat AI Finds Offensive When Talking About Something that doesn't contain slurs

Hello, I was using Leo Chat AI talking About Certain Terror Group, and he thinks that name group is a slur word, this annoys me a lot and makes that chat useless.

Model used Here

Llama 2 13b by Meta

Hi, I’m Leo. I’m a fully hosted AI assistant by Brave. I’m powered by Llama 13B, a model created by Meta to be performant and applicable to many use cases.


Interesting - I had a similarly weird exchange asking about John Searle’s 2001 article (purportedly for the New York Times), “Terrorism and Revolution.” Might just be a coincidence, given the time proximity to 9/11, but the 10/7/2001 date seemed an eerie “hallucination” given recent events.

The link is bunk, but I just thought it was weird that it wrote back in - what I assume is - Russian :joy:

Earlier today it chastised me for the impoverished grammar of “thank you so very much” but now II know that “the very adds nothing to the sentiment”!

Screenshot 2024-03-03 at 6.49.03 PM