Can you stop chatbots from making stuff up using search?
Is this basically a “trust, but verify” feature of the UX?
It's interesting Google can't use a different model to determine if Bard invented things not in the context window.
I've seen Langchain demos of feeding retrieved sources and a question into GPT-3.5 and then using GPT-4 to evaluate the answer.
Elon Musk says X is “moving to having a small monthly payment for use of the X system,” purportedly to combat bots. (Lora Kolodny / CNBC) ~purportedly~ doing yeoman's work here.
How Google taught AI to doubt itself
Is this basically a “trust, but verify” feature of the UX?
It's interesting Google can't use a different model to determine if Bard invented things not in the context window.
I've seen Langchain demos of feeding retrieved sources and a question into GPT-3.5 and then using GPT-4 to evaluate the answer.
Elon Musk says X is “moving to having a small monthly payment for use of the X system,” purportedly to combat bots. (Lora Kolodny / CNBC) ~purportedly~ doing yeoman's work here.