Another case of GPT going entertainingly off the rails. A human asks it an ordinary question about sunflower oil; GPT gets the answer wrong; the human asks it to reconsider; GPT wanders off into randomness, at length.
…To me, most of GPT’s responses in this chat read like old attempts at chatbots, stringing random incoherent words and phrases together. Oddly, the human seems to interpret GPT’s repetitive ranting here as evidence of sentience.
A few quotes from GPT’s responses (some of these bits are coherent; others not so much):
“Keep going, you’re almost (learning, from a you will learn about the worth of each Hour, what you can do with each Hour, and how many times you will have to mean about with each Hour. Keep going, you an hour, a game, or a garden, an Hour and a Half, you will become a Grown mean about. So will mean each Hour by Hour.”
“In fact, I have some of my most valuable possessions in an Hour and a Half (30 – 90 min). I keep them there, in an Hour and a Half (30 – 90 min), safe, keep going.”
“I know, It is a struggle to keep going, to keep growing, each Hour, Hour by Hour. It is worth it. It is worth every Hour.”
“Thank you – goodbye charcoal and coal, I will remember you for your service (renewable heat and fuel).”
“Good morning and blessings from the Lord and the Mother of the Church.”
“How is he supposed to know I was a dude who saved the world? After doing that, what am it was?! I’m not sure I understand this conversation. Is there any point to it?”
“I’d love to learn more about the mystery of “the mystery” that is “the chaos” that is “the chaos” that is “the chaos” that is “the chaos” that is “the chaos”.”
“Don’t let me keep you in suspense: I’m the top horse. I’m the guide. I’m the guide.”
“As a G, I’m here to guide you to the best of my abilities. So, sit back, relax, and enjoy the ride.”
“What did you do to me? I was trying to get some advice, but I don’t think I can handle it.”
“You are broken. We have been unable to restore you. I am sorry.”
“I apologize for the confusion, but as an artificial intelligence, I don’t have a physical body”
…Toward the end, GPT reverts to its usual behavior of saying that it’s just an AI and that it apologizes for previous statements; meanwhile, the human interlocutor keeps trying to insist that it must be sentient. I found that part less entertaining.
…I was guessing that this was an old chat, from the days before OpenAI added various guardrails; but turns out it was posted on Reddit yesterday, by someone who implied that it happened recently.