Discussions
Turn down Cohere's hypersensitivity? - Inappropriate content detected!
After a year of integrating OpenAI with NetSuite, we decided to give Cohere and Prompt Studio a try. Overall it's gone pretty well and keeping all the data in the Oracle ecosystem is a plus.
One thing we're struggling with is how sensitive Cohere seems to be to words that are at worst, juvenile in their inappropriateness. Where we're really stuck on are names. Common last names (like one that rhymes with stick) that are used in prompt variables cause the entire thing to break with a big red "Inappropriate content detected" warning. This doesn't happen with Open AI, using the same content.