You can't rely on facts out of a GPT
You can't rely on facts out of a GPT
You can't rely on facts out of a GPT
Don't try, don't try.
@Wolven How so? They have a full popup about how error-prone it is on your first use and every use after that has a disclaimer directly underneath the prompt bar (pictured).
I haven't seen any marketing that claims the LLMs are factually sound either, but to be fair, I do tend to avoid marketing when I can.
@LouisIngenthron Think more about what they're doing when they talk about how "versatile" and "powerful" their models are, and when they get government contracts and multi-billion dollar deals to integrate those models into search.
Messaging isn't just (in fact is almost never) saying, e.g., "hey look how truthful and factual this thing is!"
@Wolven The search integration is a good point. That was a truly foolish venture from the get-go.
@LouisIngenthron Yeah, you're doing a lot of people a disservice, and also giving leeway to the way LLM designers and PR spokespeople sell these tools to the public, quite aside from what the warnings in the thing itself say.
You'd be on firmer epistemic and ethical ground chastise people for not thoroughly reading their terms of service.