cstanhope@social.coop ("Your weary 'net denizen") wrote:
"As a large language models, We've been trained on a large body of material (such as the libraries of Yaombaaa). This training can often help us produce correct responses to queries, but we can also produce misleading or incorrect results. Ultimately, you are responsible for vetting our responses."
Gesturing in all direction with your hands towards the apparently uninhabited surroundings, "How... Uh... How am I supposed to do that out here?"
"As a large language model, I can't help you with that, but we can cite our sources on request," says lefty. "Sources should also be checked for correctness," chimes in righty.
"Great," you say in exasperation. "So... you're saying I can't really trust either of you?"
"I aim to be helpful at all times, " says righty. "I also aspire to be only a trustworthy assistant," concurs lefty.
"But neither of you can tell me what part of what you're saying is truthful?"
"That is correct," the stone faces cheerily say in unison.